Neotectonics and Paleoseismology of the Limón and Pedro...

Neotectonics and Paleoseismology of the Limón and Pedro Miguel

Faults in Panamá: Earthquake Hazard to the Panamá Canal

by Thomas Rockwell,* Eldon Gath, Tania González, Chris Madden, DanielleVerdugo, Caitlin Lippincott,† Tim Dawson,‡ Lewis A. Owen, Markus Fuchs,§

Ana Cadena, Pat Williams, Elise Weldon,|| and Pastora Franceschi

Abstract We present new geologic, tectonic geomorphic, and geochronologic dataon the slip rate, timing, and size of past surface ruptures for the right-lateral Limón andPedro Miguel faults in central Panamá. These faults are part of a system of conjugatefaults that accommodate the internal deformation of Panamá resulting from the on-going collision of Central and South America. There have been at least three surfaceruptures on the Limón fault in the past 950–1400 years, with the most recent duringthe past 365 years. Displacement in this young event is at least 1.2 m (based on tren-ching) and may be 1.6–2 m (based on small channel offsets). Awell-preserved 4.2 moffset suggests that the penultimate event also sustained significant displacement. TheHolocene slip rate has averaged about 6 mm=yr, based on a 30-m offset terrace riserincised into a 5-ka abandoned channel.

The Pedro Miguel fault has sustained three surface ruptures in the past 1600 years,the most recent being the 2 May 1621 earthquake that partially destroyed PanamáViejo. At least 2.1 m of slip occurred in this event near the Canal, with geomorphicoffsets suggesting 2.5–3 m. The historic Camino de Cruces is offset 2.8 m, indicatingmultimeter displacement over at least 20 km of fault length. Channel offsets of 100–400 m, together with a climate-induced incision model, suggest a Late Quaternary sliprate of about 5 mm=yr, which is consistent with the paleoseismic results. Comparisonof the timing of surface ruptures between the Limón and Pedro Miguel faults suggeststhat large earthquakes may rupture both faults with 2–3 m of displacement for over40 km, such as are likely in earthquakes in theM 7 range. Altogether, our observationsindicate that the Limón and Pedro Miguel faults represent a significant seismic hazardto central Panamá and, specifically, to the Canal and Panamá City.

Introduction

The geology of the Republic of Panamá in southeastCentral America records the continuing collision betweenCentral and South America (Fig. 1) (Mann and Corrigan,1990; Silver et al., 1990; Mann and Kolarsky, 1995; Coateset al., 2004). The initial collision began in the Miocene, re-sulting in the development of the east Panamá deformed belt(Coates et al., 2004) and continued with the oroclinal flexing

of the isthmus and development of the north Panamá de-formed belt in the Plio-Quaternary (Silver et al., 1990; Mannand Kolarsky, 1995), along with the closing of the seawaybetween Central and South America in the Late Pliocene(Collins et al., 1996). With a Global Positioning System(GPS)-measured rate between Central and South Americaof ∼25 mm=yr (Trenkamp et al., 2002), this ongoing defor-mation is distributed among faults throughout Panamá andwestern Colombia, resulting in internal deformation of theisthmus by folding and faulting. Many of the major faults ofPanamá have long been recognized in the geology and geo-morphology (Jones, 1950; Woodring, 1957; Stewart et al.,1980), although they were not considered Holocene (Cowan,1998, 1999; Schweig et al., 1999; Petersen et al., 2005).Among these, the Pedro Miguel, Limón, and related faultscomprise a zone that extends from the southern flank of

*Also at the Department of Geological Sciences, San Diego StateUniversity, San Diego, California 92182.

†Now at CH2M-Hill, 25 New Chardon St. Suite 300, Boston, Massachu-setts 02114.

‡Now at California Geological Survey, 345 Middlefield Rd., Menlo Park,California 94025.

§Also at the Geography Department, University of Bayreuth, D-95440Bayreuth, Germany.

||Now at the Department of Geological Sciences, University of Oregon,Eugene, Oregon 97403.

12

3

4

1

Bulletin of the Seismological Society of America, Vol. 100, No. 6, pp. –, December 2010, doi: 10.1785/0120090342

the Sierra Maestra in north-central Panamá southward forat least 40 km, crossing the Panamá Canal between theMiraflores and Pedro Miguel Locks, and extending south-ward offshore into the Gulf of Panamá (Fig. 1; EarthConsultants International, 2006).

The northern part of this zone comprises the Limónfault, which was first recognized by Jones (1950) (fault a-13in his paper); and, although he only briefly discussed thefault, he did attribute the steepest gravity anomaly gradientacross the Panamanian isthmus as being coincident with itslocation. Woodring (1957) mapped the Limón fault as a con-tinuous (albeit sinuous) strand, whereas Stewart et al. (1980)mapped the Limón fault as two unnamed, discontinuous,right-stepping traces with a normal, down-to-the-east com-ponent. In the Cowan et al. (1998) map, the Limón faultis part of an unnamed fault system (fault number PA14) thatis thought to be Quaternary in age, with an unknown rate ofslip. It is noteworthy to mention that this map includes a faultoffshore of the northern coast of Panamá, near the Panamá–Costa Rica border, that is also named the Limón fault. Thisoffshore fault, which Cowan et al. (1998) indicate last rup-tured in 1991, should not be confused with the Limón faultthat is the subject of this study.

The Pedro Miguel fault was mapped as extending onlyabout a kilometer south of the Panamá Canal (Woodring,1957; Stewart et al., 1980) with the Miocene La Boca for-mation faulted against Miocene basalt. Recent work by EarthConsultants International (2007) demonstrates that the faultcontinues much farther south, probably extending offshoreinto the Gulf of Panamá. North of the Panamá Canal, thefault is mapped to the north-northwest for about 12 km(Woodring, 1957; Stewart et al., 1980), and Earth Consul-tants International (2006) has extended the fault to the RioChagres based on mapping of lineaments, geomorphology,and limited trenching.

As part of a seismic hazard characterization for thePanamá Canal Authority’s (ACP) expansion project, we stud-ied the geologic and geomorphic expression of the PedroMiguel, Limón, and related faults, followed by an in-depthstudy into their Holocene earthquake and displacementhistory, critical factors in the design of the new locks andassociated structures. In this paper, we discuss the geologyand geomorphology of the Limón and Pedro Miguel faultsseparately, as they are distinct faults. For each, we present ageneral description of the geomorphology, followed by thepaleoseismic results, including three-dimensional (3D) tren-ching to resolve slip. We also discuss our observations thatrelate to the Late Quaternary slip rate of these faults. We con-clude with a discussion on the regional significance of thesefaults and their implications to the seismic hazard of centralPanamá, including Panamá City and the Panamá Canal.

The Limón Fault

The Limón fault extends 20–25 km south from near itsintersection with the Gatún fault to at least the Chagres River

(Fig. 1 and Fig. 2). The northern connection of the Limónfault to the Gatún fault is obscured by the water levelsat Lago Alajuela (also known as Lake Madden), complexgeological relationships, and a lack of field exposures. Thesouthern end of the Limón fault is also obscured because itlies directly within the Rio Chagres Valley. The fault is seg-mented into two distinct sections, as it makes a pronounced3-km right bend at Nuevo Vigía, before continuing north toform the western shoreline of Lago Alajuela (Fig. 2).

We first identified and mapped the geomorphologyalong the Limón fault using a digital elevation model andaerial photography as part of a study to assess the likelihood

Figure 1. Location map of central Panamá, showing the PedroMiguel (PMF), Limón (LF), Rio Gatún (RGF), Miraflores (MF),and Azuero-Sona (A-SF) faults. The inset provides the regional tec-tonic framework: NAP, North American plate; CaP, Caribbeanplate; CoP, Cocos plate; PFZ, Panamá fracture zone; NP, Nazcaplate; ECT, Ecuador–Colombian trench; SAP, South Americanplate. Arrows and circles represent the GPS velocities from theCASA campaign network relative to South America (see Trenkampet al., 2002). The color version of this figure is available only in theelectronic edition.

2 T. Rockwell et al.

of the fault’s activity and to search for potential trench sites(Fig. 2; Earth Consultants International, 2006, 2007). TheLimón fault exhibits many of the classic landforms asso-ciated with active strike-slip faulting, including deflectedand offset stream channels, beheaded channels, pressureridges, aligned notches, linear valleys, and linear sidehillbenches and troughs. Every stream that intersects the faultis right-laterally offset by varying distances, depending uponthe age and size of the stream (Fig. 3), and many of the largerstreams have a 1-m to 3-m-high knickpoint at or immediatelyupstream from the fault, reflecting either the difference inerodibility between the andesitic basalt and Tertiary sedi-ments or a small normal component of slip on the fault.In addition to the offset streams, the ground surface showsevidence of small sidehill benches where the fault crosses thefaces of slopes, a degraded moletrack from the last earth-quake, displaced terrace risers, and small-scale extensionaland compressional warping of the ground surface.

Paleoseismology and Slip Rate

We excavated two trenches across the geomorphicexpression of the Limón fault north of the Río Chagres inan area of ranching and farming. Trench T1 was excavatedacross a linear depression where we expected to have arecord of sedimentation. Trench T2 was excavated in a smalldrainage divide between two large deflected channels adja-cent to the 1.6-m deflection of a small channel, so part of thisexercise was to confirm the colocation of the fault and thedeflected channel. A series of small hand-excavated fault-parallel trenches (trenchettes) were dug within the 1.6-mdeflected channel to resolve displacement on the channel thal-weg and to provide timing of the most recent surface rupture.The timing of past surface ruptures was determined largelyfrom trench T1 and the small trenchettes in the deflection.

Trench T2 exposed the fault juxtaposing siltstone andclaystone of the Caimito formation on the east against highlyweathered and sheared andesitic basalt on the west (Fig. 4).

Figure 2.5 Map of central Panamá showing the geology and location of major faults superposed on a digital elevation model. The geologicunits were grouped by age (Basement, Paleogene, Neogene)6 and are based on the mapping of Woodring (1957) and Stewert et al. (1981). TheLimón and Pedro Miguel faults are shown, as are the Rio Gatún and Miraflores faults. Note that there are many small, unnamed faults alsomapped in the vicinity of the Canal where exposures allowed for their recognition. The location of the branch of the Camino de Cruces thatcrosses the Pedro Miguel fault is indicated. The color version of this figure is available only in the electronic edition.

Neotectonics and Paleoseismology of the Limón and Pedro Miguel Faults in Panamá 3

Figure 3. Detailed map of tectonic geomorphology near the trench site along the central Limón fault. (a) Detailed study area relativeto the Chagres River and Lago Alajuela (Madden Reservoir7 ). (b) A small rill offset, measured in the field as about 4.2 m of right-lateral offset.(c and d) Sketch of field-measured offsets and their correlation to an oblique aerial photograph. The arrows on streams in (c) indicate flowdirection; the offset hatched line denotes a ridgeline. Note the locations of the trenches in (d). The color version of this figure is available onlyin the electronic edition.


The fault exhibits a low dip of about 30° in this exposure, andsplays upward into a 2-m-wide zone of cracks in the upper1–2 m from the surface. A distinctive cobble lag composedof 2–10-cm-sized andesitic basalt clasts is preserved on eachside of the fault; this gravel is warped and faulted down intothe fault zone. The cobbles are interpreted as the result ofaxial flow along the fault as they are preserved on both sidesof the fault, and the channel margin was exposed in thistrench about 2 m west of the fault zone. We argue that thegravels are sourced from a stream located about 75 m northof T2 (see Fig. 3), as that is the first drainage that we found tobe transporting clasts of similar size and lithology. The cob-ble layer was capped by a silty fine sand deposit that is inter-preted as the result of overbank sedimentation. We collectedan optically stimulated luminescence (OSL) sample from thissilty sand layer, which yielded an age of 5:0� 0:7 ka onquartz (Table 1). Considering that the cobbles are still hardand largely unweathered and that the rate of weathering inPanamá is rapid and severe (formation of oxisols under atropical rainforest environment), we consider this as support-ing evidence that the axial fluvial deposits are Holocene inage, corroborating the OSL dating. Furthermore, as discussed

subsequently in this paper, paired radiocarbon and OSL agesfrom trench T1 yield similar dates (within a few hundredyears), further substantiating the OSL age of the axial fluvialdeposits.

Immediately north of the interfluve into which we exca-vated trench T2, there is a channel margin that slips off to thenorth in the direction of its source stream (Fig. 3). We inter-pret this channel margin as being laterally displaced by theLimón fault, with the terrace riser right-laterally offset30� 5 m. Considering that the offset terrace riser is cut intothe trenched interfluve, where the 5000-year-old streamgravel was exposed, the offset must have occurred after thestream capture that led to incision below this level. If thestream was actively flowing along the fault prior to capture,as seems reasonable, and if the terrace riser is close to the ageof the incised terrace deposits, then this implies a slip rate ofabout 6 �2:1

�1:6 mm=yr, although this rate could also be inter-preted as a minimum rate because the riser is youngerthan the actual age of the fluvial deposits preserved in theinterfluve.

Trench T1, excavated across a swale or depression atthe base of the break in slope formed by the Limón fault,

Figure 4. Log of the north face of trench T2 across the Limón fault at a small drainage divide (Fig. 3d). Note that the main faultjuxtaposes Late Oligocene Caimito formation siltstone on the hanging wall against weathered andesitic basalt on the footwall. The primaryfault was seen to dip at about 30° to the east, with a 2-m-wide zone of transtensive faulting near the fault tip. (W, west; E, east.) The colorversion of this figure is available only in the electronic edition.

Table 18 Dates and Calibrated Ages for Radiocarbon and OSL Samples Collected

from Limón Fault Exposures*

Sample NumberLaboratoryNumber Stratigraphic Unit

14C Age(Years B.P.) Calibrated Age (2σ)

Limón Trenchette 115887 surface channel 185� 35 A.D. 1640–1700 (20.8%)A.D. 1720–1820 (50.1%)A.D. 1830–1880 (5.8%)A.D. 1910–1960 (18.6%)

Limón T1-OSL-2 20b 1400� 200

Limón T1 C-5 116288 50b 970� 40 A.D. 990–1160 (95.4%)Limón T1 C-2 116036 50a 985� 35 A.D. 980–1160 (95.4%)Limón T1 C-7 116037 60b 1430� 35 A.D. 560–660 (95.4%)Limón T1-OSL-1 70 2000� 300

Limón T1-OSL-3 85 1300� 200

Limón T2-OSL-1alluvium abovebedrock in T2 5000� 700

*Calendar ages are given for the 2σ (two sigma) intercepts for the 14C dates; OSL sample ages are given as B.P.


exposed multiple colluvial deposits in fault contact that weused to reconstruct the paleoearthquake chronology alongthis segment of the Limón fault (Fig. 5). Intensely shearedand weathered andesitic basalt bedrock (unit 500) wasexposed west of the fault, whereas sheared and weatheredclayey siltstone of the Caimito formation was exposed atthe east end of the trench (unit 400; Fig. 5). These two bed-rock units are expected to be in fault contact at shallow depthbelow the base of the trench exposure, as they were in trenchT2. Near the ground surface, however, the fault is expressedas an approximately 6-m-wide zone of high-angle faultingthat is bounded on the southeast by an antithetic fault.Numerous fault splays between the antithetic and main faultdisrupt and displace colluvial units of varying ages. Thestructural relations between the main, south-dipping faultand the north-dipping antithetic fault have resulted in the for-mation of a graben along a negative flower structure that

has been episodically filled with colluvium, preserving thepaleoseismic record. We collected seven samples of detritalcharcoal from these deposits, three of which were adequatefor radiocarbon dating. We also collected three OSL samplesto help constrain the ages of the faulting events and to com-pare with the radiocarbon results (Table 1).

The colluvial deposits have been repeatedly displaced bya number of faults that are interpreted as predominantly strike-slip, forming slivers of faulted colluvium and alluvium thathave locally been faulted out of the plane of the exposure.Many of the units have clear mismatches in their thicknessacross individual fault strands, which is a clear indicator ofstrike-slip. As the deposits are primarily colluvial in nature,the fine resolution of stratigraphy and exact placement ofevent horizons was difficult to resolve. Nevertheless, we usedcolor and textural differences to identify as many as 15 dif-ferent colluvial and alluvial deposits in the graben (units

Figure 5. Log of trench T1 across the Limón fault, showing evidence for repeated Late Holocene displacements (Fig. 3d). See text fordescription and evidence of discrete events, which are indicated by letter (W, T, Q, N). Radiocarbon dates are in bold and calibrated to theAD time scale (OxCal v. 3.5, Ramsey, 2000), whereas the OSL dates are given in thousands of years before present (ka B.P.). The radiocarbondate and event pdfs are shown in the box at the lower right. Units are described as follows: 10—Modern topsoil (A horizon): silty fine tocoarse sand with scattered pebbles; dark grayish brown; organic-rich; with abundant roots. 20—Colluvial wedge: silty sand to subangularpebble gravel; reddish brown. immediately postdates event T. 30—Colluvial wedge: associated with the penultimate event (event T); gray;source material and grain size vary from clayey silt to sand to coarse, angular pebble gravel. 40—Channel and graben fill: brown to grayishbrown clay loam grading upward to organic-rich, dark grayish brown silty fine to coarse sand (A horizon); scattered pebbles. 50—Graben fill:clayey silt with pebble gravel and less than 1% volume of rounded cobbles; CaCO3 present in matrix and coating clasts. 60—Channel deposit:yellowish brown fine sand with gravel; locally clast-supported; CaCO3 coating some clasts; capped by an A soil horizon. 70—Colluvialwedge: light brownish gray clayey silt to sandy clay with coarse sand and pebbles; CaCO3 coating clasts; capped by an A soil horizon.80—Colluvium: light yellowish brown clayey silt to silty clay with sand and pebble gravel; CaCO3 present in matrix and coating clasts.90—Buried organic soil: grayish brown clay to sandy clay with scattered gravel. 95—Colluvial graben fill: grayish brown clayey silt, finesand, and coarse pebble gravel. 100,110—Undifferentiated colluvium: light grayish brown, massive clayey silt and sand to fine pebble gravel;contains dark manganese oxide nodules. 115—Colluvium: light gray clayey silt to sand with rounded pebble gravel; CaCO3 coating clasts.400—Caimito formation: clayey siltstone; gray to olive gray; highly weathered and sheared. 500—andesitic basalt (bedrock): weathered andhighly sheared; locally reduced to a fine sandstone in appearance. The color version of this figure is available only in the electronic edition.


10—115). Some of these colluvial sediments (i.e., units 20and 50; see Fig. 5) consist of reworked, weathered clasts ofandesitic basalt, whereas others appear to consist of reworkedweathered siltstone and claystone of the Caimito formationmixed in with the basalt clasts. We also used the presenceof buried soils (as are present capping units 40, 60/70, and90) to distinguish former surfaces. In the following para-graphs, we describe the evidence for each recognized surfacerupture, from the ground surface to the bottom of the trench,with the caveat that we may have missed evidence for someslip events due to the generallymassive and bioturbated natureof the colluvial sediments.

The most recent event, herein named event W, rupturesnearly to the surface on several fault strands. In fact, it prob-ably did rupture all the way to the surface through unit 10,based on the presence of a surface scarp and micromorphol-ogy associated with a degraded moletrack (rupture trace), butactive mixing processes in the surface soil, along with con-tinued colluvial deposition, likely have obscured the faulttrace in the upper 10–20 cm. Several fault strands that rup-tured in event W clearly displace the base of unit 10 and alllower units.

Charcoal was not present in units 10 or 20 to placeconstraints on the youthfulness of event W in trench T1.However, the trenchettes across the small, displaced channelabout 3 m north-northeast of trench T2 (Fig. 3) containedabundant charcoal that we use to constrain the timing of themost recent event (Fig. 6). Samples of detrital charcoal werecollected from this displaced channel for radiometric anal-ysis, which placed the age of this event as post-AD 1640(Table 1). However, the density of charcoal suggests that itoriginated from burning and clear-cutting of the forest duringestablishment of the local ranch during the nineteenthcentury. Thus, event W is historical in age, and slip on thisfault should be associated with one of the historically re-ported events, such as a poorly known earthquake in 1849that collapsed stone buildings in Panamá City, an 1855 eventthat reportedly was felt in Colón, Gatún, and Panamá City, orthe 1873 event that was also felt both in Panamá City andColón, and strongly shook the bridge over the Río Chagres.A possible fourth alternative is rupture during an aftershockof the great (M 7.9) 1882 earthquake.

To resolve slip for event W, we hand-excavated fivesmall trenches (trenchettes) in the area of the offset channel,with their axes perpendicular to the channel and parallel tothe fault. Two of these trenchettes were most helpful for thisstudy, exposing the edge of the charcoal-bearing channeldeposit that is offset by the Limón fault, as discussed in thepreceding paragraphs. Using the southwestern edge of thechannel as a piercing point, we measured a minimum of1.2 meters of brittle right-lateral slip for event Won this fault(Fig. 6). Because the trenchettes were excavated to withincentimeters of the primary fault trace, the uncertainty for thisdisplacement is small. Observations of this swale prior totrenching suggested the presence of two main fault traces:a western one offsetting the swale approximately 1.6 meters

and an eastern trace offsetting the swale an additional0.4 m, for a total displacement of about 2 m. The westernstrand with 1.6 m of apparent deflection corresponds tothe minimum of 1.2 m of brittle displacement resolved in

Figure 6. Interpreted photographs of the small, 1.6–2 m offsetswale, located a few meters north of trench T2. Several small tren-chettes (Fig. 3d) exposed a charcoal-bearing buried channel fromwhich we resolved a minimum of 1.2 m of brittle slip on one ofthe fault strands. The swale is deflected 1.6 m, with a maximumof about 2 m. The color version of this figure is available onlyin the electronic edition.

9


the trenchettes. For the eastern strand, the other trenchetteswere less definitive, as we did not observe distinct channeldeposits with clear edges that we could use as piercingpoints. Therefore, from this exercise, we determined thatthe minimum lateral displacement for event W on the Limónfault at this location is 1.2 m, although as much as 1.6–2 mare likely from the geomorphology (this would include near-field plastic deformations).

The penultimate event recognized in trench T1, desig-nated herein as event T, produced rupture on several faultstrands near the northern side of the graben that displace units40 and below and are overlain by unit 20. There are other faultstrands within the graben that displace the base of unit 40 butcould not be traced upward very far and which do not offsetthe base of unit 10. Further, a colluvial wedge (unit 30) wasapparently generated by production off a scarp from this eventon one of the main faults bounding the graben on the south. AT symbol within an octagon is used in Figure 5 to showwhereevidence for event T is fairly clear. In all, as many as five faultstrands may have ruptured in event T.

The age for event T is poorly constrained by two radio-carbon dates from unit 50. These ages are consistent andnearly identical (Table 1) at about AD 1060 (AD 980–1160at 2σ, based on our OxCal model shown in Fig. 5), and theyindicate that event Toccurred betweenAD1060 and the age ofthe channel offset in the trenchettes, dated as post-AD 1640.We also submitted OSL samples from units 20 and 70 thatshould have helped constrain the timing of this event.However, both samples yielded an age that is several hundredyears older than the radiocarbon results. We interpret theseOSL results to reflect the role of inheritance that is commonin rapidly deposited colluvial deposits that have not been com-pletely bleached prior to deposition. Additionally, the doserates obtained from the deposits were low, probably reflectingsevere leaching of the natural radiogenic minerals because ofthe high tropical precipitation, and therefore the OSL resultsshould be viewed as maximum ages due to this potential pro-blem. Nevertheless, the fact that the OSL and radiocarbonresults are similar indicates that inheritance is small. This isan important observation when considering the ∼5 ka age ofthe fluvial deposits in T2 and its implications for the slip rate.

The third oldest event interpreted from this trench expo-sure is designated as event Q and is so indicated on Figure 5.This event apparently occurred when the top of unit 70 was atthe surface on the west side of the graben and unit 60 was atthe surface on the east side. Units 60 and 70 darken to thesurface, which we infer as evidence that an A soil horizoncaps these units, indicating the occurrence of some timebetween deposition of these units and the overlying unit 50.Furthermore, this appears to have been a large event thatproduced a significant scarp that generated the depositionof units 40 and 50. Unit 40 is interpreted as the A horizondeveloped in colluvial unit 50, based on darkening of the soilmaterial upward to the top of unit 40. On the west side of thegraben, unit 50 fills against and is bounded by the fault,

whereas unit 40 is present across the fault (and was subse-quently offset in events T and W).

Several fault strands and fissures, denoted with a Q inFigure 5, apparently ruptured within the graben itself. Someof these display significant mismatches in stratigraphicthickness across them, indicative of significant lateral slip.On the east side of the graben, unit 60a was apparently back-rotated along the antithetic fault by a substantial amount, afterwhich clay-rich unit 40� 50 accumulated in the depressionproduced by this backrotation. The fault is dashed, as wecould not discern whether this was the actual slip surfaceor a degraded scarp. The observation that the main antitheticfault has not reruptured since event Q also argues that this mayhave been a larger event than either of the subsequent ruptures.

The timing of event Q is constrained by a radiocarbon agefrom unit 60 as postdating AD 560 (2σ range AD 560–660)but predating the two radiocarbon ages fromunit 50 (AD980–1160; Fig. 5). An OSL age from unit 85 (1:3� 0:2 ka)(Table 1) is consistent with these radiocarbon ages but sug-gests that the event falls toward the younger end of the pos-sible age range. Thus, event Q occurred between AD 560 and1160 but was more likely after about AD 700, based on theOSL age of unit 85 (which provides a maximum age).

Evidence for older events is also present in trench T1,but the ages of the alluvial and colluvial units are currentlyunconstrained. One event designated as event N drops thebase of unit 70 (a buried topsoil) along the western graben-bounding fault and resulted in deposition of more soil toproduce the wedge-shaped colluvial deposit of unit 70. Otherfault strands are seen to offset the base of unit 70 but not thetop. These observations indicate that the event occurredwithin unit 70 itself.

Below unit 70, we are less confident on the rupturehistory, given that weathering and repeated rupture by subse-quent earthquakes has obscured unit correlations. One possi-ble event is suggested by fault strands that offset and drop unit80 but do not break unit 70within the graben, about 0.5 m eastof the main western graben-bounding fault. Another olderevent is inferred to break units below unit 90, which is inter-preted as another buried topsoil. Unit 90 itself may have beendown-faulted at the graben margin, but that relationship isnow obscured by subsequent events. Finally, there are strati-graphic units, such as units 90 and 95, which appear to beaccumulations within the graben that could be sedimentationafter an event. However, as lateral slip is the dominant sense ofmotion, we are not confident as to the origin and extent ofthese units.

In summary, three surface ruptures have occurred on theLimón fault in the past ∼950 to 1400 years, two in the past∼370 to 950 years, and one in the past 370 years. The mostrecent event W was associated with a minimum of 1.2 m ofright-lateral slip and may have had as much as 2 m ofdisplacement. A nearby channel is deflected about 4.2 m(Fig. 3), suggesting a potentially larger amount of displace-ment in the penultimate event. Although poorly dated atpresent, the results do show recurrent Late Holocene rupture.


Calculations using the 95% confidence ranges on the intervalages of the timing of the earthquakes yields an averagerecurrence interval of 450–600 years, based on only threeevents.

The Pedro Miguel Fault

The Pedro Miguel fault has been previously mapped as astructural boundary (Fig. 2) (Woodring, 1957; Stewart et al.,1980) south and west of the Panamá Canal as the contactbetween the post-early Miocene basalt and the EarlyMioceneLa Boca formation. North of the Canal, it is mapped as form-ing the contact between the Early Miocene Las Cascadas andtheEarlyOligocene Panamá formations,while farther north, itlies entirely within the LateOligoceneCaraba formation, witha possible northern extension cutting the La Boca and LasCascadas formations.

Our geomorphic analysis suggests that the Pedro Miguelfault is longer than originally mapped by Woodring (1957).We found that the fault likely continues southward offshoreinto the Pacific Ocean where Taboga Island (see inset inFig. 2) may indicate an uplift at a left step on the fault. Tothe north, the fault extends at least to the vicinity of the townof Chilibre, making the onshore length of the fault at least30 km. North of Chilibre, the fault expression becomesweak, suggesting that it dies out, steps slip west to the south-ern extension of the Limón fault, or steps slip east to a horse-tail splay of faults south of Lake Alajuela.

The geomorphic expression of the Pedro Miguel fault isgood to fair, with this subdued signature in part due to an ab-sence of a vertical component of slip on the fault and also bythe general absence of a resistant rock (such as basalt) on onlyone side of the fault alongmost of its length. Nevertheless, thefault is expressed geomorphically by an alignment of linearvalleys, ridges, and escarpments and is generally visiblethrough the rainforest cover. Several offset and deflecteddrainages show that the fault experiences principally right-lateral motion, and this sense of motion is clear in the Cocolíarea immediately south of the Panamá Canal.

Slip Rate of the Pedro Miguel Fault

At Cocolí, the most recent channel incisions in twoparallel, east-flowing Late Quaternary drainages are bothright-laterally offset 100� 20 meters by the north-strikingPedro Miguel fault (Fig. 7). This estimate of displacementis based on reconstructions using the 1914 topographic mapsof the area that show the channel morphology prior to anysignificant earth moving in the 1940s associated with theThird Locks project (Fig. 7).

The northern drainage exhibits more deflection, about200 m, than the southern channel if a low terrace to thenorthern drainage is considered (Fig. 7). However, the south-ern drainage is underfit for its drainage area and likely isoffset from the northern drainage source, a distance of about400 m. If correct, then there should be fluvial deposits flow-

ing along the fault that are preserved in the interfluve regionbetween the two channels. To test this, we excavated a num-ber of trenches in the interfluve area to map out the fault zoneand find the fluvial gravels, if present. Figure 8 shows thelocation of the fault zone through the interfluve area betweenthe two deflected channels, as exposed in several trenches,and there are indeed fluvial gravels preserved (Fig. 9) thatreflect flow from the northern channel drainage area to theunderfit southern channel mouth. With the addition of threefault-parallel trenches, we distinguished three separate chan-nelized fluvial deposits, Qoal1, Qoal2, and Qoal3 (Fig. 8)that are all interpreted to have originated from source channelB in Figure 7.

The fault zone bounds the fluvial deposits on the west,indicating that these channels are truncated by the fault anddisplaced laterally from their source. Qoal2 and Qoal3 arebound by the primary fault, whereas Qoal1 is bound on thewest by a high-angle secondary strand, and this channel unitis found only within the fault zone (Fig. 8). The Qoal3deposit is composed of subangular clasts of basalt in amottled gray and red clayey matrix, channels Qoal1 andQoal2 contained a basal gravel composed of subroundedclasts capped by highly weathered reddened finer deposits.The channels represent flow along and east of the fault priorto the capture of the current drainage system. Thus, there hasbeen as much as 200 m of postcapture displacement on thesechannels, with as much as 100 m of post-deep incision dis-placement on both modern channels.

We attempted to date the three channel deposits withOSL because we recovered no detrital charcoal from any ofthe channel deposits. However, the quartz used in the OSLdating produced an unstable signal, which calls into questionthe validity of these OSL ages, so we did not use them forcalculation of a slip rate.

To provide some constraints on the ages of the offsetchannels and to place a first-order constraint on the longer-term slip rate of the fault, we invoke a climate-driven streamincision model. The deflected channels at Cocolí grade to theprimary drainage, the Río Cocolí, which in turn is graded tosea level. This site is only a few kilometers from the coast, so amajor drop in sea level would be expected to induce signifi-cant incision of the Río Cocolí and its primary tributariesbecause a lowered base level results in higher stream gradi-ents, which increase stream power. Considering that sea levelwas at or near its minima for the period between about 23 and15 ka (Bard et al., 1996), we infer that the modern channels(A-A’ and B-B’) correspond to incision and erosion duringthis last major eustatic sea level lowstand. We also infer thatthis is when the southern channel (channel A-A’ in Fig. 7) hadsufficient stream power to initiate significant headward ero-sion into the basalt west of the fault. Backfilling of the chan-nels during sea level rise at the close of the Pleistocene hasresulted in broad, alluvial-filled channel mouths; this is par-ticularly apparent for the southern, underfit channel (Fig. 7).

Postincision motion on the Pedro Miguel fault has re-sulted in the current configuration with about 100� 20 m of

10


offset of both channels A and B. If we are correct that theincision event corresponds to the last major eustatic sea levellow stand at 19� 4 ka, this yields a Late Quaternary slip rateof 3:5–8:0 mm=yr.

The age of the initial incision may be as much as twice asold if the incision is the result of stream capture. Stream cap-tures are more likely during periods of relatively high sealevel, as the channels are presumably graded to the higher baselevel and have the greater potential to spill across to a steeperdrainage that may be juxtaposed to the offset drainage source.With this model, the initial capture may have occurred duringthe stage 3 high-stand at ∼45 ka. If the low terrace in thenorthern drainage corresponds to this event, and we use the200� 30 m of deflection on the terrace edge as postcapturedisplacement, this suggests a similar ∼5 mm=yr rate.

Paleoseismology

We excavated several sites where sufficient stratigraphywas present to either resolve the timing of the past severalevents or to resolve displacement through 3D trenching, so

as to place constraints on the plausible size of future earth-quakes. Themain active trace of the PedroMiguel fault strikesan average of N12°W, from the Río Chagres to the southernmargin of Miraflores Lake (Fig. 2). South of the lake nearCocolí, the main fault strand is expressed as an en echelonseries of transpressive petals exploiting the west-dippingbedding planes of the La Boca formation. This main strandis approximately coincident with, though easterly of, the con-tact between the Early Miocene La Boca formation to the eastand slightly younger basalt to the west. At greater depth, thefault likely forms the La Boca formation/basalt contact.

We concentrated on the main trace, which was stronglyexpressed in the landscape morphology, as described pre-viously. Several small, 2–3-m right-lateral stream deflections(subsequently confirmed by trenching) were present alongthe main Pedro Miguel fault in the Cocolí area. One of thesewas developed into a 3D trench site (trench T21; see Fig. 8)to confirm that the deflection reflected actual displacement.Another 3D site at trench T14 (see Fig. 8) was developedadjacent to a larger channel deflection. The Pedro Miguel

Figure 7. (a) Interpreted oblique aerial photograph of the Pedro Miguel fault in the Cocolí area south of the Panamá Canal, with the areaof detailed trenching shown in Figure 8 indicated by the dashed black box. (b) The 1914 topographic map of these offsets, with theirreconstruction shown in (c) by 100 m and in (d) by 400 m. The color version of this figure is available only in the electronic edition.


Figure 8. Map of trenches that we used to map the fault and displaced channels in the interfluve area between channels A and B (Fig. 7)in the Cocolí area. Note also the locations of T14 and T21, where we conducted detailed 3D excavations to resolve displacement per event.The color version of this figure is available only in the electronic edition.


fault zone is locally broad through the Cocolí area where weconducted all of the 3D work. The primary fault at both loca-tions is low-angle, dipping about 30° to the west with multi-ple strands stepping from one bedding surface to anotherwithin the La Boca formation. Thus, the displacements de-termined from 3D trenching in this area probably reflectminimum values. Nevertheless, they provide solid informa-tion on the sense of slip, as well as timing, for past events. Inthe following section, we discuss each of the two 3D sitesseparately and then combine the interpretations into a com-mon rupture history.

Timing and Displacement of Buried Channelsat the T14 Site

Trenches T9 and T14 were excavated south of the maininterfluve area discussed previously, immediately south ofdrainage A-A’ in Figure 7, with T14 located only a coupleof meters north of T9 (Fig. 8 and Fig. 10). Both trenchesexposed the primary trace of the Pedro Miguel fault, whichin this area dips shallowly to the west, and several high-anglefaults that are transfer structures between the stepping low-angle primary faults.

Trench T9 exposed a sequence of nested gravelly chan-nel and colluvial deposits located a few meters east (down-slope) from the fault (Fig. 11a). Trench T14 (Fig. 11b)exposed the same gravel and colluvial units with much betterstratigraphic resolution. Most of these units in T14 wereobliquely overthrust by the La Boca formation bedrock, andstructural and stratigraphic relations suggested evidence ofrecurrent motion with up to three preserved events, as is dis-cussed subsequently in this paper. Also of interest was theobservation that the edge of the upper channel deposit (unit6) was closer to the fault in trench T14 than observed in

trench T9, suggesting that if this trend continued northward,the channel could intersect the fault and provide a potentialpiercing point to resolve slip. To investigate this, we exca-vated two additional trenches to the north, labeled T14a andT14b (Fig. 10 and Fig. 11c,d). As each successive cut to thenorth is closer to the active drainage, the stratigraphic sectionis expected to young to the north, a consideration that mustbe kept in mind in the ensuing discussion.

We divided the stratigraphy into ten primary units andseveral secondary units based mainly on the presence or absence of grain-supported gravel deposits, contrasts in color,obvious changes in texture, and the presence of buried topsoil/colluvial units that are interpreted to represent past groundsurfaces. Unit 1 is the modern topsoil (A horizon) and isinterpreted to be partly colluvial in nature. Unit 2 is a gravellychannel that was present above the fault in trenches T14 andT14a, with its base in fault contact, indicating that unit 2 pre-dates themost recent surface rupture. In trenchT9, this depositwas situated only on the footwall (east) side of the fault. Theserelationships indicate right-lateral strike-slip; however, be-cause the edges of the unit 2 channel had been removed duringexcavation of trenches T9 and T14,we did not try to use this asa piercing point due to the large uncertainties.

Units 3 through 5 represent a sequence of colluvial soilsand minor fine-gravel channel alluvium that, in part, fill orcap the channel deposits of unit 6. Unit 6 is subdivided intotwo subunits, 6a and 6b, with 6b representing the basalgravel and 6a representing the channel fill. Unit 6 is oneof the two channels for which displacement was resolved.Units 7 and 8 are fine-grained, organic-enriched clayey siltdeposits that are interpreted as colluvial and soil units, re-spectively, that accumulated during a period between graveldeposition. Unit 9 is the lower major gravel interpreted as the

Figure 9. Interpreted photomosaic of a portion of trench T6 (see Fig. 8 for location), showing channels Qoal1 and Qoal2. Note that inaddition to the main fault zone, numerous secondary faults break the La Boca formation, with some also displacing the paleochannel deposits.The color version of this figure is available only in the electronic edition.


axial channel deposit for the primary drainage and is the old-er of the two channels for which we resolved displacement.Unit 10 is a buried topsoil unit developed into the underlyingbedrock. This is important because one of the interpreted pa-leoearthquakes postdates unit 10 at the fault, as discussedsubsequently. In the following discussion, we refer primarilyto the channel deposits of units 6 and 9.

Age of Alluvium in the Trench T14 Site

Themany trenches at the T9 and T14 site exposed numer-ous pieces of detrital charcoal, with charcoal occurring innearly every unit. We initially selected a suite of 12 samplesfrom trench T14 to run for radiocarbon dating that includedcharcoal frommost units. All of the dates from this initial suiteyielded ages in the range of 1400–1600 radiocarbon yearsbefore present (ybp) (Table 2). These dates are in the sameage range as four samples that we acquired from unit 5 intrench T9 and indicate that either the entire section was de-posited in a very short period of time or that therewas a sourceof carbon dating to that period that was incorporated into eachof the units. A likely source of such carbonwould be the burn-ing of the forest by the indigenous people of the area to makeway for agriculture or the intense occupation of a site upslopefrom trench T14 in the 1400–1600-ybp time frame.

To test whether the charcoal has been reworked into thedeposits or whether the entire section was deposited over200 years, we submitted a larger number of samples from the

T14 site to see if young charcoal is mixed in with the 1400–1600-ybp group. The results in Table 2 show that all samplesrecovered from this site yield similar ages, although theyoungest date (T14a C-14) was recovered from the youngestunit (unit 2). We then ran two mean residency time (MRT)radiocarbon ages from the buried A horizons to test whetherthe soil ages were consistent with the detrital charcoal ages,which would indicate an older source for the detritalcharcoal. In general, the MRT dates yielded slightly youngerages than the detrital charcoal but are still consistent with thedetrital charcoal dates. These data suggest that the detritalcharcoal ages are, in fact, a generally valid representationof the ages of the units.

Taking the youngest dates from each stratigraphic unitand assuming these are maximum ages for the host unitsbecause they are all detrital charcoal, unit 2 is younger thanAD 671–872, unit 3 is younger than AD 566–655, unit 7 isyounger than AD 467–650, unit 8 is younger thanAD 594–688, unit 9 is younger than AD 432–600, and unit10d is younger than AD 396–547. The humic date from theunit 5 soil yielded an age range of AD 660–875, consistentwith some residence age for the detrital charcoal but also ar-guing that the entire section is indeed about 1400–1600 radio-carbon years old.

Of the two OSL samples from trench T14a, one ofthem (OSL-T14a-1) is in complete agreement with the radio-carbon dating results, whereas the second one (OSL-T14a-2)

T9 - North Face

cros

s-tr

ench

-1

T14a North Face

T14b South Face

T14 North Face

0 1 2 3 4 5m

Exploratory trench faces

~2.1 m,unit 6

~4.5 m minimum,unit 9

South edge of Unit 6 feeder channel

South margin of channel 6

South margin of channel 9

2-5

2-4

2-32-1

2-21-5

1-4

1-3

1-2

1-1

N

Figure 10. The T14 3D trenches (see Fig. 8). The southern margin of unit 6 gravel is indicated on both sides of the fault, whereas thesouthern margin of the unit 9 gravel is truncated at the eastern fault strand. The locations of the individual slices shown in Figure 13 areindicated. Slip on the unit 6 channel is measured in the slip direction, as the fault expresses oblique slip at this site. Slip for the unit 9 channelis a minimum, as the location of the feeder channel is unknown but must be no farther south than the unit 6 feeder channel west of the fault.The color version of this figure is available only in the electronic edition.

11

12

14


yielded a date roughly twice as old as the radiocarbon age forthe same unit. Apparently in this young channel, the partialbleaching of the quartz measured in the OSL dating mayoverestimate the OSL age by as much as 2000 years, whichis inadequate for resolving the timing of past ruptures.Moreover, quartz grains were rare in the samples; and,because of the quartz stability problems we encountered withthe older OSL dates, we did not use them to determine the

ages of past ruptures, although they generally confirm theyouthful nature of the section.

Identification of Past Surface Ruptures

We identified evidence for three separate surface rup-tures that occurred in the past ∼1600 years. We refer to theseas events 1 through 3, with event 1 being the youngest.

Figure 11. Logs of trenches T9, T14, T14a, and T14b (Fig. 8 and Fig. 10). Units are described in the text, as is the evidence for eachfaulting events. The locations of the slices (Fig. 13) are indicated where there intercept the faces of trenches T14a and T14b. The color versionof this figure is available only in the electronic edition.


Table 213 Radiocarbon Dates and Their Calibrated and OSL Ages from Trenches at Cocoli on the Pedro Miguel Fault

Sample ID* Unit ID† CAMS# (LLNL Lab Number) d13C‡ 14C Age‡ Calibrated Age (2σ)§

Trench T9 samplesPM-T9-1 5 129344 �27:44 1415� 30 585 AD (95.4%) 663 ADPM-T9-2 5 129345 �26:17 1530� 30 432 AD (95.4%) 600 ADPM-T9-3 5 129346 �26:57 1635� 30 343 AD (95.4%) 534 ADPM-T9-4 5 129347 �31:53 1445� 35 558 AD (95.4%) 655 AD

Trench T14, samples T14a and T14bT14a-14 2 133758 �25 1255� 40 671 AD (95.4%) 872 ADPM-T14-20 3a 130850 �25:30 1440� 30 566 AD (95.4%) 655 ADT14-21 3a 134291 �25:04 1485� 25 467 AD (95.4%) 645 ADT14a-3 upper 3 133750 �27:94 1505� 45 424 AD (95.4%) 616 ADT14a-13 middle 3 133757 �25 1485� 35 443 AD (95.4%) 649 ADT14a-4 3 133751 �21:46 > Modern modernPM-T14-5 lower 3b 130767 �24:92 1625� 30 353 AD (95.4%) 536 ADT14a-7 lower 3 133753 �26:82 1505� 35 430 AD (95.4%) 617 ADT14b-10 clay blob 133768 �25:87 1470� 30 535 AD (95.4%) 652 ADT14b-11 clay blob 133769 �22:60 1565� 30 427 AD (95.4%) 593 ADT14b-3 ? 4=5 133761 �25 1455� 35 551 AD (95.4%) 651 ADT14-3 upper 5 134294 �26:49 1405� 30 572 AD (95.4%) 663 ADT14-3 ha upper 5 134498 �25 1270� 50 660 AD (95.4% 875 ADPM-T14-19 upper 5 130852 �26:12 1535� 30 419 AD (95.4%) 574 ADT14-18 upper 5 134290 �29:19 1530� 35 417 AD (95.4%) 570 ADT14a-5 5 133752 �25 1705� 40 245 AD (95.4%) 416 ADT14a-8 5 133754 �25 1690� 40 249 AD (95.4%) 426 ADT14b MRT-1 5 133861 �25 1480� 60 433 AD (95.4%) 656 ADT14b MKT2 5 134288 �25 1435� 35 563 AD (95.4%) 658 ADPM-T14-14 lower 5 130851 �25 1605� 35 388 AD (95.4%) 544 ADT14-15 ha 5 to 7 contact 134500 �25 1420� 30 442 AD (95.4%) 766 ADT14b-6 base 5 133764 �26:86 1485� 35 434 AD (95.4%) 644 ADT14b-4 top 6/base 5 133762 �25:11 1785� 30 132 AD (95.4%) 338 ADT14b-5 6b 133763 �25 2520� 130 970 BC (95.4%) 366 BCPM-T14-2 upper 7a 130853 �26:18 1495� 20 540 AD (95.4%) 616 ADPM-T14-6 7a 130854 �27:54 1525� 30 432 AD (95.4%) 604 ADT14-23 ha 7 134501 �25 1480� 35 467 AD (95.4%) 650 ADT14-23 7 134292 �26:90 1465� 30 467 AD (95.4%) 650 ADT14a-2 7 133749 �27:08 1470� 30 441 AD (95.4%) 646 ADT14a-9 7 133755 �27:34 1465� 35 438 AD (95.4%) 650 ADT14b-7 7 133765 �27:98 1460� 30 443 AD (95.4) 649 ADT14b-8 7 133766 �25 1510� 35 434 AD (95.4%) 635 ADT14a-10 7 133756 �28:00 1535� 30 419 AD (95.4%) 574 ADT14a-12 7a/7b 133774 �24:35 1945� 30 37 BC (95.4%) 130 ADPM-T14-10 7b 130855 �24:55 1645� 35 263 AD (95.4%) 534 ADPM-T14-12 top 8 130857 �28:77 1485� 30 455 AD (95.4%) 647 ADPM-T14-4 8 130856 �26:06 1670� 30 258 AD (95.4%) 430 ADT14-1 8 134296 �27:97 1390� 25 584 AD (95.4%) 663 ADT14-1 ha 8 134497 �25 1700� 35 253 AD (95.4%) 415 ADT14-7 lower 8 134295 �25:97 1510� 35 432 AD (95.4%) 623 ADT14-7 ha lower 8 134499 �25 1385� 35 594 AD (95.4%) 688 ADT14-11 lower 8 134293 �21:49 1435� 30 580 AD (95.4%) 670 ADPM-T14-13 9a 130858 �25 22060� 1100 Not calibratedPM-T14-9 9a 130859 �13:00 1555� 30 424 AD (95.4%) 584 ADT14a-1 top 9 gravel 133748 �23:44 1735� 35 242 AD (95.4%) 405 ADT14a-15 9 gravel 133759 �27:03 1645� 25 258 AD (95.4%) 504 ADPM-T14-8 9 gravel 130860 �25:91 1530� 25 432 AD (95.4%) 600 ADPM-T14-16 10b 130861 �25:00 1595� 35 396 AD (95.4%) 547 AD

T14 OSL SamplesOSL-T14a-1 2 1240� 110

OSL-T14a-2 lower 3 3310� 270

Trench T21 sampleT21-LWC-1 133860 �25 175� 35 1660 AD (95.4%) 1954

(continued)


For this discussion, we refer to the main low-angle fault asfault 1 (Fig. 11). In the footwall of fault 1 are several higher-angle faults that we interpret to be transfer structures betweensurficial, low-angle fault petals, and these are collectivelyreferred to as fault 2 (as a zone).

Event 1. Event 1 is obvious in all exposures and places theLa Boca formation bedrock over soil and alluvium alongfault 1. The fault can be traced in most exposures up into thesurface soil units, appears to truncate the base of unit 2 intrenches T14 and T14a (Fig. 11b,c), and resulted in deposi-tion of colluvium in the upper part of unit 3 (in this case, it isclear that the upper part of unit 3 must be younger than unit 2,but unit 3 was locally massive and a distinct event horizonwas not possible to distinguish, so we could not delineate thepre-event and postevent components of unit 3 in most expo-sures away from fault 1).

In the T14 trenches, the age data only limit the occur-rence of event 1 to younger than about 1300 years ago (sam-ple 14 from unit 2 in T14a: AD 671–872), though it is likelymuch younger because the fault is well expressed up throughthe soil. Nevertheless, from the T14 trenches, the age dataprovide only a maximum date for this event as depositionhad ceased at this site.

Event 2. Event 2 was initially interpreted in trench T14from rupture of footwall fault 2, which displaces units7–10 but could not be traced above the base of unit 7(Fig. 11b). Furthermore, unit 7 itself appears to be a colluvialwedge complex that was generated by slip on fault 1, as thedeposits thin and pinch away from the fault. We also debatedwhether the upper part of unit 7 represents a separatecolluvial wedge and another event, although the subsequentexposure in trench T14a appears to have clarified this issue.

Trench T14a was cut only 50–70 cm north of the T14exposure but revealed substantially enhanced clarity in theoccurrence of event 2 (Fig. 11c). Fault 2 moved in this eventand involved all units up through unit 7. The fault wasobserved to roll almost flat, folding unit 7 as a recumbentstructure onto the gravel of unit 9, and this fold was subse-quently buried by the undeformed gravel and colluvium of

units 6 and 5. As is discussed further, we resolved displace-ment of unit 9 on fault 2 only; and, based on the generationof colluvium from fault 1 in this event, there was almostcertainly movement on both faults in event 2.

Evidence for event 2 is also seen in trench T14b(Fig. 11d), where faulted gravels of unit 9 are overlain byunfaulted gravel of unit 6 along fault 2. Because most ofthe section between units 6 and 9 is eroded by the unit 6gravel, this trench did not provide as good a resolution onthe precise stratigraphic position of event 2, although it doessupport its occurrence.

The deformation associated with Event 2 is large andalmost certainly coseismic. The event placed rock and soila substantial distance over the unit 9 gravel, offset unit 9a significant distance (as discussed subsequently), and is bur-ied by undeformed units. These observations all argue thatthe deformation is not the result of creep but is due to slipin a very short period of time, as in an earthquake.

The event horizon for event 2 is bracketed by units 6(AD 556–655) and unit 7 (AD 467–650), indicating thatthe age of event 2 is in the time frame of AD 640 (see moredetailed discussion on the ages in the next section).

Event 3. Event 3 is inferred from the T14 exposure wherethe oldest part of the section is preserved. At the base of thesection, unit 10d is a dark, organic-rich buried soil developedinto the La Boca formation bedrock. A single piece of detritalcharcoal was recovered from unit 10d on the east side of thefault in trench T14 that indicates this unit was buried no morethan about 1600 years ago. This appears to have been theground surface prior to deposition of the overlying section.At fault 1, a triangular-shaped block of probable La Bocabedrock (unit 10c in T14, Fig. 11b) is juxtaposed over unit10d and is, in turn, overlain by the wedge-shaped colluviumof units 10a and 10b. Event 3 is interpreted to postdate unit10d, resulting in placement of rock on soil and subsequentdeposition of a fault-derived colluvial wedge. Evidence forevent 3 was not exposed in trench T14a, in spite of the closeproximity of the two trenches (50–70 cm), as the gravel ofunit 9 had apparently cut out the older section of unit 10.

Table 2 (Continued)Sample ID* Unit ID† CAMS# (LLNL Lab Number) d13C‡ 14C Age‡ Calibrated Age (2σ)§

Trench T33 samplesT33-MRT-4 Sta. 5.5 (�28–40 cm) 134299 �25 425� 35 1418 AD–1620 AD

T33-MRT-6 Sta. 6.9 134301 �17:28 2925� 25 1126 BC–926 BC

T21 OSL samplesT21-HE1 ditch 2440� 220

T21-BE Ch. A 8520� 820

T21-IE1 3D Ch. C/D 6480� 710

T21-IE2 3D Ch. C/D 7050� 710

*Samples with names that include T9 or T14 are from the T14 3D site. Ha refers to humic acid dates.†The unit designations correspond to the logs of T9 and T14.‡Some samples were run for 13C, and the radiocarbon ages have been corrected accordingly.§The calibrated ages are reported as the entire 2σ range and were calibrated in OxCAL (Bronk-Ramsey, 2005).


In trench T14 (Fig. 11b), the precise relationshipbetween event 3 and unit 9 is not clear. It is plausible that unit9 predates the deposition of the colluvial wedge associatedwith event 3, as the colluvium of unit 10 and the gravel of unit9 are at essentially the same level. If unit 9 is actually olderthan the upper part of unit 10, then our numbering system is abit awry. The observation in T14 is that the unit 9 gravel endsat the toe of the unit 10 colluvium, and unit 7 overlies both.This is important when discussing the displacement of unit 9and whether it is from one or two events. From our existingexposures, this relationship is not clear.

The timing of this event is bracketed by units 10d and10a, but there are no age data for the colluvial wedge of unit10a. Thus, we use the age data from unit 9 (AD 432–600) toplace an upper bound constraint on the age of event 3 andthat from 10d (AD 396–547) for the lower bound. Thus,these data place the event age at around AD 455.

To better constrain the event ages, we take the youngestdates from each unit and construct probability distributionsfor the event ages in OxCal (Ramsey, 2000; Bronk-Ramsey,2001; Bronk-Ramsey, 2005), resulting in surface rupturesaround AD 455, AD 640, and after AD 900 (Fig. 12). Usingdata from other trenches, the most recent event (MRE) mustbe historical (as discussed subsequently in this paper and in

Earth Consultants International, 2007) and is likely the largehistorical earthquake of 1621. From this, and assuming thatwe did not miss an event, it is evident that earthquake recur-rence has been irregular.

In summary, it is likely that there have been three surfaceruptures on the Pedro Miguel fault in the past 1600 years,with the earlier two events well dated and the age of the mostrecent event poorly constrained from these exposures. Weargue from dating from another trench (T33; Earth Consul-tants International, 2007), from the historical record, andfrom offset of the Camino de Cruces that event 1 is likelythe large historical earthquake of 1621 that strongly damagedthe original Panamá City (Panamá Viejo). If correct, then allthree events occurred in a ∼1200-year interval, which arguesfor a fairly short recurrence interval of about 600 years forthe past 1600 years or so. These data also argue for irregularrecurrence. We discuss the importance of this in the interpre-tation of the slip rate after we present evidence for slipper event.

Resolution of Slip

As part of this study, we tested the possibility of a pre-served subsurface offset of channel units 6 and 9 by first

400BC 200BC BC/AD 200AD 400AD 600AD 800AD 1000AD

Calendar date

Sampled 3 390AD (95.4%) 560AD

0.0

0.2

0.4

0.6

0.8

Rel

ativ

e pr

obab

ility

AD 455

0.0

0.2

0.4

0.6

0.8

200BC BC/AD 200AD 400AD 600AD 800AD 1000AD 1200AD

Calendar date

Sampled 2 600AD (95.4%) 680AD

Rel

ativ

e pr

obab

ility

AD 640

1000BC 500BC BC/AD 500AD 1000AD 1500AD 2000AD

Sequence PanamaBrief {A= 8.6%(A'c= 60.0%)}

C_Date 1855 100.0%

Event 1

T14a-14 98.1%

T14a-3 2.1%

T14-3ha 24.0%

T14-15ha 115.0%

Event 2

T14-23 63.3%

T14-23ha 115.6%

T14a-2 75.9%

T14b-7 96.5%

T14-1 61.2%

T14-7ha 24.4%

T14-11 26.6%

T14-9 83.3%

T14-8 94.2%

Event 3

T14-16 81.7%

AD 1621 EQ

AD 455

AD 640

Railroad built

Figure 12. Probability distributions of the youngest radiocarbon dates from each unit. The event ages are plotted individually. Note thatdeposition at the T14 site apparently ended around 1200 years ago, so the most recent event(s) is poorly constrained in age. Atmospheric datafrom Reimer et al. (2004)15 ; OxCal v 3.10,Bronk-Ramsey (2005). The color version of this figure is available only in the electronic edition.


excavating T14a about 0.5 m farther north from the T14north wall with the intent to simply provide a fresh exposure(Fig. 10). The actual distance between T14 and T14a variedfrom about 50 cm in the western part of the new excavationto about 70 cm where the channels were exposed east of thefault. Surprisingly, the edge of the unit 9 channel had movedall the way to fault 2, where the channel gravel was now infault contact (Fig. 11c), indicating that the gravel intersectedthe fault (the piercing point) on the east side of the faultsomewhere between the two cuts. Although we inadvertentlycut this piercing point out, we know its location to be withinthat 70-cm interval. For this exercise, we place the piercingcontact for the unit 9 channel margin as halfway between thetwo walls and assign a�35 cm uncertainty to the location ofthe piercing point.

We then excavated trench T14b to explore the distribu-tion of the channel deposits ∼3 m to the north. This trenchexposed La Boca bedrock juxtaposed over the unit 6 channelgravel along fault 1, with the unit 9 gravel in fault contactwith fault 2. Thus, the eastern piercing point for the unit6 channel was present somewhere between the trench facesof T14a and T14b. The unit 9 channel was found in faultcontact in both exposures, so it must be offset more thanthe 3-m distance between the trench faces.

West of the fault, the south margin of the unit 6 gravelwas exposed in the west end of trench T14b, and the gravelfilled the north wall of T14b. (This relationship can be seenin Fig. 13, cut 2-1 which looks west—the feeder channel islocated north of the north wall of trench face 14b, as mappedin Fig. 10). The location of the edge of the gravel was sur-veyed, and its position is known to within a half meter ofwhere it intersects fault 1 on the west side of the fault. Thus,tracing the south margin of channel unit 6 into the fault fromthe east provides slip on fault 1 in the MRE, as fault 2 did notmove in the MRE. Further, it provides a piercing point for theminimum displacement of channel unit 9, as the location ofthe unit 9 feeder channel cannot be any farther south than theunit 6 feeder channel because a fault-parallel trench we ex-cavated on the west side of the fault did not expose anothersource for the gravel.

We then excavated a series of slices parallel to the fault totrace the margin of the unit 6 channel into the fault and to ex-plore the distribution of other units. In all, ten slices weremade (Fig. 10 and Fig. 13; eight interpreted photomosaic logsfor cuts 1-1 through 2-5 are included in Fig. 13) that allowedfor precise location of the unit 6 channel edge. These cuts alsodemonstrated that units 5 and 6a are the upper part of the samechannel fill sequence for which unit 6b is the basal gravel.

For this exercise, we use the gravel of unit 6b as the pri-mary piercing point, but we also note that the channel marginis seen to be offset a similar amount. Strike-slip was mea-sured parallel to the average strike of the fault of N12E—not parallel to the low-angle faults that expressed obliqueslip—because the direction of motion is oblique in the planeof the low-angle fault and there is a dip component of slip.Unit 6b was found to be offset about 2.1 m across fault 1. In

cut 1-3 (Fig. 13), the top of the channel fill expresses a right-separation of about 1.8 m, and in cut 1-4 (Fig. 13), the rightseparation is as much as 2.4 m. Taken together, and consider-ing that our uncertainty on the precise location of the gravelchannel west of the fault is about a half meter, we assign2:1� 0:5 m for the minimum slip in the MRE. It is a mini-mum because we cannot discount the probability that somewarping occurred on the hanging wall on this low-angle faultor that other vertical strands (similar to fault 2, which weinterpret as a step-over transfer structure) in the hanging wallmay have accrued some slip. In trenches T9 and T14, wenoted several minor faults in the hanging wall that displacedthe topsoil, supporting this argument.

The additional cuts confirmed the continuous presenceof the unit 9 gravel in contact with fault 2 between the facesof T14a and T14b, yielding an absolute minimum displace-ment of 3 m, none of which occurred in the MRE becausefault 2 did not slip in the MRE. If the feeder channel contin-ued along its trend across fault 1 to fault 2 (an additional 5 mat the elevation of the unit 9 channel gravel), then we inferabout 4.5 m of post-unit 9 displacement (Fig. 10). An alter-native method is to sum the minimum slip values from Fault1 in the MRE (2:1� 0:5 m) and fault 2 (minimum of3:0� m) in the penultimate event, in which case the unit9 channel is offset at least 5:1� 0:85 m. However, basedon the observation in trenches T14 and T14a that fault 1 alsolikely moved when the unit 9 channel was displaced (pre-sence of a colluvial wedge shed from fault 1 in the penulti-mate event), this displacement value is a gross minimum.Thus, after deposition of unit 9, there has been at least∼5 m of lateral slip, with the likelihood that the actual dis-placement is larger. Because of the uncertainty between theage of deposition of unit 9 and the occurrence of event 3, itis plausible that this large displacement occurred in threeevents. However, from the T14 exposures, we cannotpreclude that the displacement in the penultimate event(event 2) was substantially larger than that which occurredin the MRE (event 1).

In summary, we have resolved a minimum displacementof 2:1� 0:5 m for the gravel of unit 6, which can only beattributed to event 1. For unit 9, we have resolved a minimumadditional displacement of 3 m, which is likely the result ofevent 2. As discussed previously in this paper, there was alsoadditional slip on fault 1 during event 2 that resulted in thecolluvial wedge of unit 7. Thus, it appears that event 2 waslarger than event 1. However, because of the uncertain rela-tionship in the age of units 10a and 9, we cannot precludethat some of the displacement associated with unit 9 occurredduring event 3.

Timing and Displacement of Buried Channelsat the Trench T21 Site

The excavation of trench T21 (Fig. 8) initially was con-ducted to resolve the location of the Pedro Miguel fault andto test whether a ∼2:5–3-m deflection of a surface channel

16


Figure 13. Slices of the block of soil between trenches T14a and T14b. The location of the slices are shown on Figures 10 and 11. Notethe location of the unit 6 and 9 gravels as each slice is cut progressively into and across the faults. The main, low-angle fault is seen as thenearly horizontal bold line across each of the exposures. The color version of this figure is available only in the electronic edition.


was potentially fault related. Unfortunately, this particulargeomorphic feature was inadvertently destroyed during grad-ing for vegetation removal before we could get it surveyed,but it was very similar to the 2.6–3�-m stream-wall offsetthat is still visible in the adjacent drainage, located 50 mnorth of trench T21 (Earth Consultants International, 2007).

Trench T21 exposed a channel filled with a distinct,light-colored gravel offset by a west-dipping, low-angle fault(Fig. 14). The distribution of gravel exposed in the trenchwalls, herein named channel A, required the presence of lat-eral slip, confirming the geomorphic offset and making thisan attractive site to explore the kinematics of the fault in threedimensions.

Therefore, after initial logging of trench T21, we cut anumber of new trenches oriented more or less parallel tothe fault zone and perpendicular to the original trench T21(labeled trenches T21A through T21L on Fig. 15). In partic-ular, we were interested in resolving the lateral distribution ofthe channel A gravel, exposing other channels that, ifpresent, might provide useful information on the fault’sQuaternary displacement and improve our understanding thethree-dimensional character of the fault zone.

Channel A and the Feeder Channel

West of the fault, we exposed the feeder channel in threetrenches (T21A, T21B, and T21C), which, taken together,show that the channel bends to the right (south) as it ap-proaches the fault in the far field. Two meters west of thefault, the channel bends back to being nearly perpendicularto the fault, which is opposite the sense expected for lateraldrag. The feeder channel is not only the source for channel A

but also for several other channels exposed in the lateraltrenches, as discussed subsequently.

East of the fault, the margins of channel Awere exposedin a sequence of trenches excavated parallel to the fault(Fig. 15), with a small, 4-inch-wide (10.2-cm) slot trenchat the fault demonstrating the precise location of the southernchannel margin, which we use as a piercing point. However,the southern margin of the feeder channel on the west side ofthe fault was cut out in the initial excavation of T21, so thereis an uncertainty equal to the width of the inner trench cut ofT21 (∼1 m) as to the precise amount of displacement on thispiercing point.

Figure 15 shows the distribution of the channel A gravel(including the feeder channel gravel) as it approaches andcrosses the fault. The northern margin is well preservedand is offset about 1.7 m. However, it would have beenexpected that this margin would have been trimmed byadditional flow if the channel was active at the time ofdisplacement. The southern margin is precisely known eastof the fault, but to the west it was cut out in the initial excava-tion of T21, as previouslymentioned. For the southernmarginof the feeder channel, we place the location in the middle ofT21 and assign an uncertainty that is equal to half thewidth ofthe inner trench slot cut (about 1 m). Using this location, weestimate the southernmargin to be offset 2:2� 1 macross themain fault, similar to the northern margin.

In addition to themain fault trace, these trenches exposedseveral secondary faults that, as in the T14 area, seem to betransferring slip between the low-angle primary structures. Atleast one of these hanging-wall faults cuts the feeder channelgravels and part of the deflection west of the fault on thehanging wall could be attributed to lateral warping, sodisplacement along the primary fault must be assumed to

Figure 14. Annotated photomosaic of the south wall of trench T21 (prior to excavation of the 3D trenches; see Fig. 8 for location).Channel A is exposed in the trench face only east of the fault, making this interesting for a 3D excavation. The color version of this figure isavailable only in the electronic edition.

17


be a minimum. The channel wall of a small gully, locatedabout 25 m north of trench T21, was measured to be offset2.6–2.8 m at the fault, and this likely provides a reasonableestimate of the actual displacement in the most recent event.

Offset of Channels B, C, and D

South of trench T21, the fault-parallel trenches on the eastside of the fault exposed a complex of nested channels south ofchannel A, which we herein name channels B, C, and D,

AwAe

Bw Be

Cw CeDw De

Ew Ee

Fw

Fe

GwGe

HeHw

IeIw JeJw

KeKw

LeLw

T-21

Continuation of T-21

~8 m

Iw

4" Slot Trench

Feeder Channel

A

B

C

D E

F

G

H

I J

K L

~2.2 m

N

0 2 4

8 10m

6

channels B, C and D

channel A

Feeder Channel

~1.7 m

Area of detailin Figure 17

channel E

channel F

Figure 15. Map of 3D trenches at the T21 site (see Fig. 8), along with generalized locations of major channel elements. The ancillarytrenches are labeled A through L, with the east and west walls noted by the lower case w or e. Note that the trench walls were sloped, so thereis an upper and lower edge indicated for each trench. The color version of this figure is available only in the electronic edition.


respectively, in succession southward. The channels are bestseen as separate features in the logs of lateral trenches I and J(Fig. 16) but are grouped together in Figure 15. In allexposures where each channel is preserved, the stratigraphicrelationships between the channels show that channel D isoldest and channel B is youngest. Furthermore, there is ageneral pattern of lowering of the thalweg elevation fromthe oldest (D) to the youngest (B) channel, indicating long-term incision during the cutting and accumulation of thedeposits associated with these channel forms.

Channels D and C have gravelly sand at their bases, withscattered cobbles up to 15 cm in size. The gravelly sand layeris generally capped by fine-grained alluvium or colluvium,some of which displays bedding. In contrast, channel B isfilled with massive clayey silt and generally lacks gravel, in-dicating a change in hydraulic conditions between the fillingof channels C and B. Channel A, as discussed earlier, is grav-elly sand and represents a return to conditions inferred to havebeen similar to the cutting and filling of channels D and C.

We traced each of the channels and their deposits fromtrench face to trench face (Fig. 16 and Fig. 17). In thedetailed map of the trenches and channels (Fig. 17), thetrenches are indicated by an inner (base) and outer (top) linebecause most of the trench faces had some slope and becausethe surveyors tended to survey the base of the trench some-what inside the actual base of the trench face. The criticalchannel contacts, however, were directly surveyed, so theyappear about midway between the top and bottom of thesurveyed trench locations because the channels were locatedmidway up the trench faces.

As seen in Figure 16 and Figure 17, the B/C/D channelcomplex is abruptly truncated by the fault zone, and there isno equivalent channel complex preserved in trench H west ofthe fault. We did, however, observe another channel in trenchH located south of the projection of the B/C/D channel com-plex, and there was concern that perhaps this channel (chan-nel E, Fig. 15) somehow fed to the channel complex east ofthe fault. To test this, we excavated a number of destructivecuts through the block between trenches H and I and deter-mined that (1) the B/C/D channel complex is abruptly trun-cated at a high-angle fault that roots into the low-angle mainfault exposed in trench T21 and (2) there are no gravel orother channel deposits obscured in that block by whichthe channel E could have fed to the B/C/D channel complexwithout left-lateral slip. Because we are certain of the direc-tion of slip on the youngest channel (right-lateral), we areconfident that this possibility is highly unlikely and we dis-count it. From this exploration, we infer that channels D, C,and B all originated from the same feeder channel as channelA, although not necessarily with the same width and depthconfiguration as channel A.

Figure 18 shows a series of three reconstructions thatassume all of the channel deposits originated from the feederchannel. In the first reconstruction, we back-slip the fault andalign channel A directly in front of the feeder channel, usingan average of 1.9 m based on the 1.7 and 2.1 m offsets

resolved from the north and south channel margins, respec-tively. As previously discussed, this 1.9-m reconstructionprobably underestimates the actual displacement becausethe southern margin is slipped into a protective positionand should be better preserved, whereas the northern marginis subject to continued erosion. Thus, the southern marginreconstruction of about 2.1 m (with the attendant uncertain-ties) is a better estimate.

For channel B, we realign its southern margin to resolveabout 4.6 m of displacement. The northern margin of channelB has been cut out by channel A, so this is the only estimateof displacement available to us. (Note that we are correlatingto the closest channel margin and not to the preserved depos-its because the feeder channel is filled with gravel that isclearly younger than the mud that fills channel B. Neverthe-less, subtracting the 1.7–3.6 m displacement of channel Afrom the ∼4:6 m displacement of channel B results in a rea-sonable minimum estimate of displacement on the main faultfor the penultimate event of about 1–3 m.

Reconstruction of channels C and D requires additionalinterpretation. It is uncertain as to whether channels C and Drepresent a channel complex that was offset by a singleearlier event (event 3, counting back from the youngestdisplacement event that offsets channel A) or whether eachchannel represents offset in two separate discrete events(events 3 and 4). The problem is twofold. First, the stratigra-phy in all exposures demonstrates that channel C is cut intoand is slightly lower than channel D and is therefore younger.It is also closer to the feeder channel, so it can be inferred torepresent an additional event. However, channels D and Care, on their own, considerably narrower than the feederchannel; and, if they represent individual events, then thefeeder channel itself must have grown in width in the timesince the cutting of channel D. This is possible, and we notethat channel D is both the narrowest and topographicallyhighest of the channels, whereas channel A is the lowest(or nearly so) and the widest. This could represent the growthof a rill into a small channel over time, during which as manyas four earthquakes produced slip on the Pedro Miguel fault.Alternatively, channels C and D are collectively as broad asthe feeder channel and could represent a nested channelsequence that was offset from a similar-sized feeder channel.In this scenario, there have been only three events that haveoccurred since the cutting of channel D.

The primary difference in interpretation reflects theinferred size of event 3 (or events 3 and 4). In panel C ofFigure 18, we reconstruct channels D and C to the feederchannel to resolve about 8.1 m of displacement. We chosethis representation because these two channels are collec-tively about the same size as the feeder channel. If thisrepresentation is correct, then there have been only threeevents, and the third event back was apparently larger, withnearly 3.5 m (8.1–4.6) of slip on the main fault. Alterna-tively, if there have actually been four events since the de-position of channel D, then the average displacement perevent for all four earthquakes is about 2 m, similar to that

18

19


Figure 16. Logs of ancillary trenches H through L at the T21 site (Fig. 15). Note the locations of channel elements B, C, and D,which are used to map out their distribution and resolve lateral displacement. The color version of this figure is available only in the electronicedition.


resolved for channel A, and argues for fairly characteristicslip. The actual displacement across the entire fault zoneis likely more, based on our geomorphic observations ofsmall channel offsets in the Cocolí area, as well as the ob-servation that other secondary faults displace the feederchannel deposits in the hanging wall of the fault.

Other Channels

As noted previously, channel E is present west of thefault but has no match east of the fault that we identified.It is likely that the offset portion of channel E is located farthersouth along the fault, but we did not explore for this becausewe found no charcoal in this channel. Finally, there is acolluvium-filled channel farther south that we investigated(channel F in Figure 15). The channel is not faulted inany exposure, and the channel form appears to trend acrossthe fault with no offset. In fact, this channel is a narrow, deeprill filled with colluvium and may in fact be a historical

drainage ditch. In any case, we could find no charcoal in thischannel, but it was sampled for OSL, which yielded a ∼2:5 kadate, as is discussed subsequently. Presumably, the channel isvery young and postdates the MRE in spite of its apparentOSL age.

Age Control

Age control for the trench T21 stratigraphy is very poor.We found only three pieces of detrital charcoal, and two ofthese did not produce sufficient carbon to yield results. Thethird sample that did yield a date was interpreted to be detritalcharcoal and was recovered from the channel C alluvium intowhich channel A is incised. Thus, it must predate the mostrecent event. This sample (T21-Lw C-1) yielded a date of175� 35, which dendro-corrects to a calendar date that isyounger than AD 1660. This is problematic from two per-spectives. First, it makes the most recent event on the PedroMiguel fault not only historical but too young to correspond

Cw Ce Dw De

EwEe

Fw

Fe

Gw Ge

Hw

KeKw LeLwHe Iw IeJw Je

Be

4" Slot Trench

T

T

channelriser

channelmargin

~2.1 m

AwTrenchette designation,wall (w-west, e-east)

Fault, dashed where inactive

N

Feeder

Channel

Flow ~1.7 m

0 5m

20

Channel D

Channel C

Channel B

Channel A

Figure 17. Detail of the distribution of gravels and channels at the T21 site. The displacement for each channel element is discussed inthe text. The color version of this figure is available only in the electronic edition.

20


to the large 1621 earthquake. Other more recent earthquakesthat are reported in the literature include a poorly documen-ted earthquake on 11 January 1849 that reportedly collapsedstone buildings in Panamá City; an earthquake on 25 April

1855 that was felt in Panamá City, Gatún, and Aspinwall(Colón), and the 13 October 1873 earthquake that was feltboth in Panamá City and Colón. Though improbable,we cannot preclude that this event was associated with an

N

0 5m

T

MDT

Feeder

Channel

Flow

20

~4.6 m

~1.9 m

T

MDT

20

T

MDT

channel margin

Feeder

Channel

Flow

20

~8.1 m

Reconstruction of the channel A

by 1.9 m. Note that the southern

channel wall is the protected

margin, whereas the northern

margin would be expected to be

trimmed after offset if the channel

still sustained flow.

Reconstruction of the southern

edge of channel B, the "mud-filled

channel", to the southern margin

of the feeder channel.

Reconstruction of the gravel in

channels C and D to the feeder

channel. This reconstruction

assumes that both channels are

of similar age and were offset

in the same event. If channel

D represents offset in four

events, then an additional

reconstruction is required.

(a)

(b)

(c)

channel riser

channel riser

Feeder

Channel

Flow

channel riser

Channel BnnCC aaCC ananCCCCChhhhaaCChahannnnnnCC nneellnne BBeel BBll BBee BBBBeeeeee

Channel D

Channel C

Channel D

channel marginChannel D

Channel B

Channel A

Channel B

Channel D

Channel C

channel margin

Figure 18. Reconstructions of offset channels at the T21 site. See text for discussion on these reconstructions. The color version of thisfigure is available only in the electronic edition.


aftershock to the great 1882 earthquake that is thought tohave occurred on the north Panamá deformed belt.

Alternatively, we may have misidentified the charcoal inthe field such that what we dated was actually a burned orcarbonized root. Because we have only a single date, thisis a very difficult issue to resolve without further work anddating. To address this, we collected OSL samples from thesediments of channels A and D, as well as from the younger,colluvium-filled channel that is not apparently displaced bythe fault. The sample from the young colluvium in channelE (T21-HE1), which is likely historical andmay fill a drainageditch, yielded anOSL age of 2:44� 0:22 ka. The sample fromchannel A (T21-BE) yielded an age of 8:52� 0:82 ka, whichtaken at face valuewould suggest only a single surface rupturein over 8000 years; this is known to be incorrect from the eventchronology recorded in the T14 site, as discussed earlier inthis paper. Finally, two samples were recovered from thechannel C/D deposits (samples T21-IE1 3D and T21 IE23D) that yielded ages of 6:48� 0:71 ka and 7:05� 0:71 ka,respectively. Based on stratigraphic relationships, the depositsthat yielded these two samples must be older than the depositsthat fill channel A, which yielded the ∼8:5 ka date. From theknown relative ages, it appears that there are substantialbleaching problems in the rare quartz in the OSL samples thatwe collected from the T21 site, making their use problematicin precise dating of the sediments in these trenches.

Two last pieces of age evidence are noteworthy from theT21 complex: none of the deposits are very weathered, andthere are no saprolitic soils developed in any of the alluvialunits. Based on visual comparison to the soils at the T14 site,where we have abundant age control, the T21 soils looksimilar and are probably in the same general age range. Incontrast along the Gatún fault, the soil developed in theradiocarbon-dated 2.5–3.3-ka alluvium (Earth ConsultantsInternational, 2005) appears stronger, with a reddenedargillic horizon. From this general comparison, we infer allof the deposits in channels A through D to be Late Holocenein age and likely younger than 2.5 ka.

Offset of Camino de Cruces

The Camino de Cruces is a 1.5-m-wide cobblestone(paved) road that was first established circa 1533 to enablelarge mule pack trains to move Incan gold across the isthmus(Fig. 2). The original track extended from the old town site ofPanamá Viejo (near modern Panamá City) on the Pacificcoast, across the drainage divide, to the town of Cruceson the Chagres River, approximately across from the presenttown of Gamboa (Fig. 2). It was in nearly continuous useuntil completion of the railroad in 1855, after which the trailwas abandoned.

During the course of our investigations, several observa-tions suggested that the most recent rupture on the PedroMiguel fault may have been fairly recent and possibly his-torical in age. In addition to the problematic very young ageof carbon in faulted alluvium from the T21 offset channels, a

radiocarbon date on charcoal from a soil beneath bedrock inT33 (Fig. 8) indicated faulting in the past several hundredyears. Furthermore, several small channel deflections thatwe interpret as rill offsets appear fresh and young, with ap-parent displacements in the 2.5–3-m range. Based on theseobservations, we (T. Rockwell, and E. Gath, together with W.Lettis, and D. Ostenaa) walked the reach of the Camino deCruces near the northern mapped trace of the fault to searchfor its intersection with the fault (see Fig. 2 for general loca-tion). Using a handheld GPS receiver, we followed the trail tothe vicinity of the fault crossing as mapped by Stewart et al.(1980) and found an apparent ∼3 m right-lateral offset of thetrail and an associated low scarp across the lowest fluvialterrace near the confluence of two small drainages. Alignedwith this offset, the modern stream channel walls and theriser above the low terrace are also deflected about 3 m,whereas the modern stream channel itself is diverted right-laterally over 30 m, forming a horseshoe-shaped meander.Other geomorphic features typically associated with activestrike-slip faults were observed to the north, on trend withthe fault (Earth Consultants International, 2008).

ACP’s Topographic, Hydrographic, and CartographicSection of the Engineering Division conducted a detailedtopographic survey of the area (Fig. 19), which allowedus to analyze the geomorphic characteristics of the siteand quantify the trail deflection. In addition to topography,the survey team located the edges of the deflected streams,the thalwegs of deflected rills, and several other importantgeomorphic features that were identified by our group. Theyalso surveyed the location of hundreds of clasts along the oldroadbed of the Camino de Cruces and the three trenches thatwe had hand-excavated across the fault zone. This allowed usto map out and reconstruct the trail to resolve displacementacross the fault.

The three small trenches and two hand-cleared stream-cut exposures into the low terrace to the small creek exposedthe fault in Late Holocene alluvium (Fig. 19 and Fig. 20).The trenches confirmed that the fault is spatially associatedwith the trail deflection and the various observed geomorphicoffsets. We infer the alluvium to be Late Holocene becauseit expresses a low degree of weathering (essentially none)similar to other Late Holocene sites that we directly datedby radiocarbon on detrital charcoal. However, we found nocharcoal in any of the trenches, so we have no direct dates onthe low terrace itself.

The simplest reconstruction is to find the amount ofretrodeformation that best matches all of the deflected andoffset features (Fig. 21). The trail itself curves as it crossesthe fault, in part because the trail hugs the edge of the lowterrace south of the fault but aligns with a constructed cut inrock as the road climbs out of the terrace to the west of thefault. Thus, the overall trail deflection is greater than theactual trail offset. However, the configuration of the originalroadbed has a very sharp deflection at the fault and it ishighly unlikely that the road was built that way, consideringthe volume of traffic. The trail appears to have been rebuilt or

21


surveyed north margin of trail

surveyed south margin of trail

Scarp Trench

linear scarpand offset riser

n

Cruces trail

10

11

Constructed

retaining wall

(rutted into rock)

Cruces trail

nto rock )

rp Trench

Terrace Trench

W

2

1

N

Elevation in meters

Contour Interval = 10 cm

108

108

10911

1071

Southern Bank Exposure

fault-line inflection

modified scarp

109

111

115

?

?

?

104104

101

103

Terrace edgebase of riser

linear ridge

channel walls

deflected stream northern bank exposure

106

111111

101100

rill

deflected

Pedro Miguel fault

10 15 20 25 30 35 40 50

surveyed cobbles (to scale in cm)

0 10 20 meters

101

102

10510

3

107

113

103

102

103

102 101

106

105100

Figure 19. Annotated topographic map of the Camino de Cruces trail crossing the Pedro Miguel fault. Note the several offset rills,channel margins, and riser along with the offset of the trail itself. Note also the location of the trench and stream bank exposures usedto confirm the location of the main fault. The cobbles of the trail were individually surveyed and categorized by size. The color versionof this figure is available only in the electronic edition.


straightened across the fault after the offset because wemapped remnants of the trail divergence east of the fault,with one trail truncated at the fault and the other aligned withthe trail portion heading for the bedrock notch. Consideringall of the deflected and offset features, a reconstruction ofabout 2.8 m realigns the four well-defined piercing lines,suggesting this is the actual displacement from the mostrecent event. We assign an uncertainty of about 0.5 m to thisoffset as the overall misfit to the four features.

These observations require the PedroMiguel fault to haveruptured historically, between the trail’s completion around1533 and its abandonment in 1855. There is only one well-documented earthquake in this time frame, the 2 May 1621earthquake, that appears to have been sufficiently large tohave produce 2.8 m of displacement, and we thus attributethe offset of the Camino de Cruces to this earthquake.

The 1621 Earthquake

Based on Salcedo’s 1640 description of the main shock(Salcedo, 1640; Peraldo and Montero, 1999), shaking was sointense in Panamá Viejo that it was difficult to maintainequilibrium. Strong aftershocks continued throughout theevening and for at least 24–48 hours afterwards, with fourto five aftershocks felt every hour. Peraldo and Montero(1999) interpret Salcedo’s description of the stronger of theseto suggest that the seismic waves were observable at theground surface. For at least 15 days following the main-shock, there were three to four aftershocks every day, sopeople feared entering their houses. Aftershocks continued,although diminished in number and strength, for three-and-a-half months, until 21 August. The earthquake resulted inextensive damage or collapse of the stone and adobe struc-tures in Panamá Viejo, which is consistent with a modifiedMercalli intensity of at least VIII. Based on the damagealone, reported from a single locality, it is not possible toestimate earthquake magnitude, but the reported damageis compatible with an earthquake close toM 7, which is alsoconsistent with the observed offsets at Camino de Cruzes.

Discussion

Our paleoseismic work on the Limón and Pedro Miguelfaults demonstrates that these are active, seismically capablefaults that have ruptured during the historical period. Theyhave a relatively short return period for large earthquakes,with displacements in the range of 1.5–3 m. Consequently,they represent a major potential seismic hazard to centralPanama and specifically to Panama City and the PanamaCanal. Nevertheless, there are several outstanding issues thatwarrant discussion as they relate to the potential size of earth-quakes that may occur on this fault zone.

Foremost is whether the Limón and Pedro Miguel faultsrepresent discrete seismogenic sources or whether they typi-cally rupture in unison. At the surface, these two fault strandsare not colinear zones but rather have different strikes andnear-surface geometries (Fig. 2). As discussed earlier, theLimón fault is mapped as two separate strands separatedby a right step (Fig. 2). The right-step along the Limón faultoccurs within a topographically elevated section, whereas aright step should be a zone of extension. This observationsuggests that the apparent right-step at the surface doesnot reflect extension across a right step at depth but maybe more related to the shallow near-surface dip of the fault,as exposed in our trenches (Fig. 4). Although we have nosubsurface structural information on the northern strand ofthe Limón fault, the surface trace is linear, suggesting a fairlysteep dip. It is possible that the low easterly dip of the south-ern strand merges easterly with a high-angle northern strandat depth, allowing for a more linear fault at depth. Probablyonly detailed microearthquake studies along the Limón faultcan resolve whether this is true.

If the Limón fault does have a steeper dip at depth, withthe deep, seismogenic portion of the southern trace of thefault zone more to the east than its surface trace, this couldexplain why the Limón and Pedro Miguel faults have prob-ably ruptured together in some past events. The PedroMiguel fault was nearly vertical in all exposures north ofthe Panama Canal. This configuration would make the over-all zone considerably straighter than the surface traces of thefaults would imply. We suggest that the southern part of theLimón fault has inherited its geometry from an earlier historyof normal faulting. The Limón fault may have been reacti-vated in the Quaternary as a strike-slip fault, with its surfacegeometry heavily influenced by the earlier movement historyand consequent fault geometry.

The slip rate of this zone is also a topic that requiressome discussion. The Mid-Holocene to present rate on theLimón fault is estimated at about 6 mm=yr from offset ofa terrace riser inset into a 5000-year-old terrace remnant,although this age is dependent on a single OSL date. The LateHolocene rupture history may argue for a lower rate of about3–4 mm=yr (1.5–2-m displacements every 400–600 years),although the record of only three events is likely too short toassess a rate. In both cases, the Limón fault is a moderately

alluvium

saprolite

A horizon (topsoil)alluviumgranular

coarse-grainedbasalt

massive agglomeratic tuff

shear zone

low scarp

layeredagglomeratictuff

W ESouthern Stream Bank ExposurePedro Miguel Fault, Camino de Cruces

approximately 1 m

Figure 20. Trench log of the hand-excavated southern streambank exposure across the Pedro Miguel fault near the Camino deCruces trail. See location of this exposure on Figure 18. The colorversion of this figure is available only in the electronic edition.

22


11

Cruces trail

2

1

N

Elevation in meters

Contour Interval = 10 cm

108

108

10911

1071

?

is realigned

deflected rill

Miguel fault

10 15 20 25 30 35 40 50

surveyed cobbles (to scale in cm)

2.8 m

surveyed north margin of trail

surveyed south margin of trail

Cruces trail

10

(rutted into rock)nto rock )

W

109

111

113

1155

?100

103

channel walls

realign

deflected stream

101100

Pedro

?

estimated location of the original (pre-1621) Camino de Cruces

0 10 20

terrace riser realigns

channel margins realign

101

102

101

102

106

105

104

103

102

107

10510

3

103

Figure 21. Reconstruction of the topographic map of Figure 19 to realign rills, channel walls, and the major stream deflection. Thisrequired 2.8 m of back-slippage. Note that the Camino de Cruces still has a deflection in its alignment, which presumably allowed for theburro trains to transition from the steep trail notch cut into rock to the west as the trail transitioned onto the edge of the low terrace east of thefault. The trail deflection is now a symmetric double bend, as opposed to a strong kink associated with the original trail cobbles. The colorversion of this figure is available only in the electronic edition.


fast fault in the Holocene, which is what drives the currentseismic hazard associated with this structure.

For the Pedro Miguel fault, we used a speculativeclimate-driven incisionmodel to date the clearly defined chan-nel offsets and resolve a long-term rate of about 5 mm=yr.However, there are no reliable direct ages on any of the largeoffset features, so this rate requires further assessment.

A short-term rate can be derived from the trenchingobservations along the Pedro Miguel fault at Cocolí. Forthe T14 site, we consider two alternative interpretations.We consider the possibility that the displacement amountswere similar for all events (i.e., characteristic slip), whichimplies that the unit 9 gravels are offset by all three events.Alternatively, if unit 9 was offset in only the past two events,the penultimate event was larger than the MRE (event 1).Here, we discuss each scenario, along with their inferenceson rate and assuming that the short-term rate adequatelyreflects the long-term rate.

For one model, we assume characteristic slip—that is,each of the past slip events had similar displacements atCocolí. For the age model, we take the youngest dates fromeach unit and construct probability distributions for the eventages, resulting in surface ruptures around AD 455, AD 720,and theMRE poorly constrained to post-AD 900 (see Fig. 22).Using the observation of the offset Camino de Cruces, theMRE must be historical and is likely the large historical earth-quake of AD 1621. From this, and assuming that we did notmiss an event, it is evident that earthquake recurrence hasbeen irregular, with interval recurrence between events ofabout 265 and 920 years. If representative of the long term,this suggests a return period of about 600� 330 years.However, because the unit 6 gravel channel clearly truncatesand erodes all earlier units and because slip on unit 9 is aminimum, it is possible that another event occurred betweendeposition of units 7 and 6 for which we have no preservedevidence in the T14 site. If this is the case, then we are likelyunderestimating the rate and overestimating the return peri-od, and the Pedro Miguel fault could be quasi-episodic in itsbehavior. However, because we have no observations thatsupport such a model, we do not discuss it further.

Taking the timing of the three known events and assum-ing characteristic behavior of 2–3 m per slip event for thePedro Miguel fault, we estimate an average slip rate of about4:5 mm=yr with an uncertainty of about 50%. Consideringthe elapsed time since the MRE, it is unreasonable to arguefor increasing the rate substantially. Nevertheless, this short-term rate is similar to that inferred from interpretation of thegeomorphology and incision history.

An alternative model is to assume that the penultimateevent was large and that slip on the unit 9 channel reflectsjust the past two events. This is based on the displacement ofunit 9 along fault 2 of at least 3 m, with additional significantslip likely occurring on fault 1. Also, the MRE produced slipon only fault 1 at this site, whereas the penultimate eventproduced slip on both faults 1 and 2. This observation aloneargues that the MRE was smaller than the penultimate event.

For this scenario, we used about 5 m of slip in the penulti-mate event (≥ 3 m on fault 2 and 2 m on fault 1, similar to theMRE). Again, using the average timing between events and adisplacement value of 3–5 m suggests a slip rate of about6:7 mm=yr (3–5 m on average every 600 years), and it couldbe larger if event 3 was also large. However, we believe thatthis provides a reasonable upper bound to the slip rate on thePedro Miguel fault if the past 1600 years are a reasonablerepresentation of the long-term average.

In summary for the Pedro Miguel slip rate, the LateHolocene rupture history is consistent with a slip rate inthe half centimeter per year range, but, because the eventshave been fairly irregular in their timing and because the

Limón

Fault

500 1000 1500 2000 AD0

Limón TrenchetteLimón Event W

Limón Event T

Limón T1-C2

Limón T1-C7

Limón Event Q

~AD 455

~AD 720

T14a-14

T14a-3

T14-15ha

T14-3ha

PMF Event 1

PMF Event 2

T14-23

T14-23haT14a-2

T14b-7T14-1T14-7ha

T14-11

T14-9

T14-8

PMF Event 3T14-16

Pedro

Miguel

Fault

AD 1621 AD 1873?

AD 1621 EQ

Railroad

built - 1855

Limón T1-C5Limón T1-C1

Figure 22. Comparison of dates of surface ruptures from theLimón and Pedro Miguel paleoseismic sites. Note that event Qat the Limón fault has a very similar age to event 2 on the PedroMiguel fault, suggesting that these two faults may have rupturedtogether in that event. Also note that the Limón fault sustainedtwo ruptures in the past 1 ka, whereas we only recognize event1 for the Pedro Miguel fault. Finally, note that event 3 on the PedroMiguel fault is older than the dated record for the Limón fault, butwe speculate that it may correspond to the undated event N at theLimón site. The color version of this figure is available only in theelectronic edition.


displacement per event may have varied between 2.5 and> 5 m, there is a large uncertainty associated with theseestimates. We consider the past 1600 years to be consistentwith a best-estimated rate in the 4–7 mm=yr range, although,it is possible that the past 1600 years is an insufficient lengthof time to estimate a slip rate for this zone.

Figure 22 also plots the probability distributions for theevent ages for the Limón and Pedro Miguel faults. The oldestevent on the Pedro Miguel fault at approximately AD 455 isolder than any of the events recorded for the Limón fault.However, the penultimate Pedro Miguel event and the thirdLimón fault event have very similar ages at approximatelyAD 700 and may represent rupture of the entire onshorezone. The MRE for the Pedro Miguel has a large uncertaintyin its age when inferred strictly from the radiocarbon dating,but the offset of the Camino de Cruces shows that it is almostcertainly the historical earthquake of AD 1621. Similarly, thepenultimate Limón event is poorly constrained but overlapswith the penultimate Pedro Miguel event, suggesting thatthey could have ruptured together. In contrast, the MREfor the Limón fault is certainly younger than the 1621 earth-quake and likely represents rupture of only the Limón fault inthe late 1800s. Taken together, the data allow the Limón andPedro Miguel faults to fail together but also indicate that theLimón fault may fail independently from the Pedro Miguelfault, as may have occurred in 1873 or during an aftershockof 1882.

Seismic Hazard Considerations for Central Panamá

The Panamá Canal system and Panamá City are at sig-nificant risk from a large earthquake on either the Limón–Pedro Miguel or the Río Gatún faults. The Pedro Miguelfault crosses the Panamá Canal between the Pedro Migueland Miraflores locks and, based on extensive work onshorenear the canal and in Miraflores Lake itself, no existingCanal structures appear to directly overlie the fault (EarthConsultants International, 2007; Technos, Inc., 2006). Thus,the hazard from the Pedro Miguel fault is largely one of shak-ing and its consequent effects. As Panamá City lies only afew kilometers from the Pedro Miguel fault, renewed activityon this fault could cause substantial damage to structures thatwere not designed for strong shaking.

Implications for Regional Tectonics

The ∼5 mm=yr slip rate for the dextral Pedro Miguelfault, along with a similar sinistral rate for the Rio Gatún fault(Earth Consultants International, 2005), indicate that centralPanamá is internally deforming at a significant rate. Thisobservation invalidates the assumption of Trenkamp et al.(2002), which treated Panamá as a stable block. The basisfor this inference was that the regional GPS rates observedfrom the CASA network generally indicated rates in themultiple cm=year range, whereas the uncertainties from thiscampaign network were about a cm=yr. Although the GPSdata did suggest some internal deformation of Panamá, the

rates were generally lower than the uncertainties, so theywereignored. A first-order conclusion from our work is that therates are sufficiently high that we believe they should beconsidered in local and regional tectonic modeling.

Rockwell et al. (2010, in press) developed a new blockmodel for the deformation of central Panamá, where pureshear strain resulting from the collision of Central and SouthAmerica is partly accommodated by the conjugate faults incentral Panamá, as well as by folding and thrusting in easternPanamá and the northward oroclinal flexing of Panamá intothe Caribbean. In their reevaluation of the CASA data, theyfound that western Panamá is converging with Panamá Cityat about 8 mm=yr, consistent with an active Pedro Miguelfault. An interesting conclusion of the block model is thatthe Pedro Miguel fault is largely driven by continued short-ening in westernmost Panamá and not by the northwardoroclinal flexing of the peninsula (Rockwell et al., 2010,in press) which results in left-lateral slip in the Darien regionto the east (Coates et al., 2004).

Conclusions

The Pedro Miguel and Limón faults comprise poten-tially hazardous seismic sources to central Panamá, with arate of right-lateral faulting estimated at 4–7 mm=yr. Boththe Pedro Miguel and Limón faults have been historicallyactive, with multimeter offsets of small channels, rills, andthe historical Camino de Cruces. Although the fault passesthrough the Panamá Canal between the Pedro Miguel andMiraflores locks, missing all critical structures, its locationand rate of activity is a factor in expansion of the PanamáCanal and its new lock system. The close proximity ofPanamá City to this active fault zone and the lack of consid-eration of earthquake loads in structural design codes makethis a particularly hazardous condition should the fault rerup-ture before the current building stock is replaced withstronger, more earthquake-resistant construction.

Data and Resources

All data used in this paper were generated by us duringthe seismic hazard assessment of the Panamá Canal. Addi-tional trench logs included in the referenced Earth Consul-tants International, Inc. consulting reports are available fromthe senior author. Radiocarbon dates were generated by theCenter for Accelerator Mass Spectrometry facility at Lawr-ence Livermore National Laboratory, and optically stimu-lated luminescence (OSL) results were generated from theOSL laboratory at Cincinnati University.

Acknowledgments

The work reflected in this paper was the result of many people’s hardwork and support over several years. We especially want to thank the En-gineering Division of the Panamá Canal Authority (ACP), Luis Alfaro andMaximillian DuPuy, and the many geologists and engineers in the groupwho all rendered valuable assistance and advice. Many geologists fromEarth Consultants International (ECI) also participated in various ways,

23

24

25


and we thank them and ECI for their contributions. We thank Hugh Cowanand James Lienkaemper for excellent reviews that led to substantial im-provements in the presentation of this paper. We also benefited from theextensive field and report reviews conducted by the ACP’s Geotechnical,Structural, Seismic, and Paleoseismic Advisory Boards and Peer Reviewers,but the final interpretations and conclusions in this paper are ours.

References

Bard, E., B. Hamelin, M. Arnold, L. Montaggioni, G. Cabioch, G. Faure,and F. Rougerie (1996). Deglacial sea-level record from Tahaiti coralsand the timing of global meltwater discharge, Nature 382, 241–244.

Bronk-Ramsey, C. (2001). Development of the radiocarbon program OxCal,Radiocarbon 43, no. 2A, 355–363.

Bronk-Ramsey, C. (2005). Radiocarbon program OxCal version 3.10.Coates, A. G., L. S. Collins, M. Aubry, and W. A. Berggren (2004). The

geology of the Darien, Panamá, and the Late Miocene–Pliocene colli-sion of the Panamá arc with northwestern South America, Geol. Soc.Am. Bull. 116, 1327–1344.

Collins, L. S., A. G. Coates, W. A. Berggren, M. Aubry, and J. Zhang (1996).The Late Miocene Panamá isthmian strait, Geology 24, no. 8, 687–690.

Cowan, H. (1999). Design earthquakes for the southeast area of the canalbasin, Panamá, Consulting report to the Authoridad Canal de Panamá.

Cowan, H., R. L. Dart, and M. N. Machette (1998). Map and Database ofQuaternary faults and folds of Panama and its offshore regions: A pro-ject of the International Lithosphere Program Task Group II-2, MajorActive Faults of the World, map accompanying the U.S. GeologicalSurvey Open File Report 98-779, September 30, 2005.

Earth Consultants International, Inc. (ECI) (2005). Paleoseismic investiga-tion of the Gatún and Limón faults in Central Panamá, ECI ProjectNo. 2505; consulting report prepared for the Autoridad del Canal dePanamá, Project No. 2505, 30 September 2005, 61 p.

Earth Consultants International (ECI) (2006). Geomorphic evaluation of theMiraflores, PedroMiguel, Azota, and Caballo faults, Central Panamá,consulting report prepared for the Autoridad del Canal de Panamá,Project No. 2614.01, November 2006, 36 p.

Earth Consultants International (ECI) (2007). Paleoseismic trenching of thePedro Miguel fault in Cocolí, located immediately southwest of thePanamá Canal, Panamá, consulting report prepared for the Autoridaddel Canal de Panamá, Project No. 2614.02, February 2007, 31 p.

Earth Consultants International (ECI) (2008). Quantitative characterizationof the Pedro Miguel fault, determination of recency of activity on theMiraflores fault, and detailed mapping of the active faults through theproposed Borinquén Dam location, consulting report prepared for theAutoridad del Canal de Panamá, Project No. 2708.05, 31 January2008, 91 p.

Jones, S. M. (1950). Geology of Gatún Lake and vicinity, Panamá, Bull.Seismol. Soc. Am. 61, 893–922.

Mann, P., and J. Corrigan (1990). Model for Late Neogene deformation inPanamá, Geology 18, 558–562.

Mann, P., and R. A. Kolarsky (1995). East Panamá deformed belt: Structure,age, and neotectonic development, in P. Mann (Editor), Geologic andTectonic Development of the Caribbean Plate Boundary in SouthCentral America, Spec. Pap. Geol. Soc. Am. 295, 111–130.

Peraldo, G., and W. Montero (1999). Sismología histórica de AméricaCentral, Publicación No. 513, Instituto Panamericano de Geografíae Historia, México, D.F., 347 p.

Petersen, M., E. Schweig, C. Mueller, S. Harmsen, and A. Frankel (2005).Preliminary update of the probabilistic seismic hazard analysis forsites along the Panamá Canal Zone, 25 p.

Ramsey, C. B. (2000). OxCal Program Ver. 3.5, Radiocarbon AcceleratorUnit, University of Oxford, Available at: http://units.ox.ac.uk/departments/rlaha/orau/06_01.htm.

Rockwell, T., R. Bennett, E. Gath, and P. Francesci (2010). A revised tec-tonic model for the internal deformation of Panama, accepted toTectonics, in press.

Salcedo, Requejo J. (1640). Relación histórica y geográfica de la Provinciade Panamá, in Relaciones Históricas y Geográficas de America Cen-tral, Tomo VIII, Librería General de Victoriano Suarez, Madrid, 1908,311 p. (Historical and geographical report of the Province of Panama:1947 translation by Alice E. Westman.)

Schweig, E., H. Cowan, J. Gomberg, T. Pratt, and A. TenBrink (1999). De-sign earthquakes for the evaluation of seismic hazard at Gatún Damand vicinity, Report to the Panamá Canal Commission for InteragencyAgreement Number CNP-93786-NN-29 between the Panamá Canaland the U.S. Geological Survey (9 August 1999), 60 p.

Silver, E. A., D. L. Reed, J. E. Tagudin, and D. J. Heil (1990). Implicationsof the north and south Panamá thrust belts for the origin of the Panamáorocline, Tectonics 9, no. 2, 261–281.

Stewart, R. H., J. L. Stewart, and W. P. Woodring (1980). Geologic map ofthe Panamá Canal and vicinity, Republic of Panamá, U.S. Geol. Surv.Misc. Invest. Series, Map I-1232, scale: 1:100,000.

Technos Inc. (2006). Final Report Geophysical Investigations for the ThirdSet of Locks Project, Panama Canal, Panama, consulting report pre-pared for the Panama Canal Authority, Project No. 06-150, 22 Septem-ber 2006, 120 p.

Trenkamp, R., J. N. Kellogg, J. T. Freymueller, and H. P. Mora (2002). Wideplate margin deformation, southern Central America and northwesternSouth America, CASA GPS observations, J. S. Am. Earth Sci. 15,157–171.

URS Inc. (2008). Final report, task A: Development of design earthquakeground motions, consulting report prepared for the Panama CanalAuthority, Contract No. CMC-172538, February 2008, 292 p.

Woodring, W. P. (1957). Geology and paleontology of Canal Zone and ad-joining parts of Panama,U.S. Geol. Surv, Profess. Paper 306-A, 145 p.

Earth Consultants International, Inc.1642 East 4th StreetSanta Ana, California 92701

(T.R., E.G., T.G., C.M., D.V., C.L., T.D., A.C., E.W.)

Department of GeologyUniversity of CincinnatiCincinnati, Ohio 45221

(L.A.O., M.F.)

Geological SciencesSan Diego State UniversitySan Diego, California 92182

(P.W.)

Autoridad del Canal de PanamáBalboa, Republic of Panamá

(P.F.)

Manuscript received 27 October 2009

26

27

28

2930

31

32

33

34

35

36

37

38

39


http://units.ox.ac.uk/departments/rlaha/orau/06_01.htm






Queries

1. BSSA only allows one primary affiliation per author; all others are given in the author footnotes. (1) For Rockwell:please let me know if it should be the Earth Consultants International, Inc. in the footnote. (2) For Fuchs: please let meknow if it should be the University of Cincinnati in the footnote.

2. Please note that I replaced “San Ana” in all author addresses with “Santa Ana” to match the Earth Consultants Inter-national, Inc. address available on the Internet. Please let me know if this is NOT correct.

3. (1) Please review the rewording of the sentence beginning with “Displacement in this young event...” (done to aidnon-English speakers) for accuracy. (2) Also, I have replaced “as” with “because” in many cases since this use of “as”seems to be problematic for many non-English speakers. Please keep this in mind when reviewing the entire article tobe sure I have not altered your intended meaning.

4. Please note that I have tried to follow BSSA style conventions in the capitalization (or lack thereof) of geologicalterms.

5. Please review all figures carefully for legibility. Many of the figures appear to have broken type that may not re-produce well. For any that do not look clear in the proof, please provide a higher resolution files.

6. There is no Stewert et al. 1981 in the References—please provide complete information to be added to Reference list.(Note: there is a Stewart et al. 1980.)

7. In the first paragraph of the section “The Limón Fault,” Lago Alajuela is named as Lake Madden, but here it is givenas Madden Reservoir. All synonyms may be given with first use, but which one should be used for consistencythroughout the rest of the paper?

8. Please review material from Table 1 title that has been moved to footnotes and to column heads for accuracy. (BSSArequires short table titles.)

9. Throughout the article, trench numbers were variously referred to both with and without hyphens (e.g., T-9 and T9). Istandardized to the nonhyphenated form within the text proper to match the majority of the figures; however, pleasenote that Figures 8, 14, 15, and 16 use hyphens and that Figure 5 uses a hyphen at the top left but no hyphens at thebottom right. You may wish to consider submitting revised figures with the hyphens removed for greater consistencywithin your paper.

10. Would it be correct to change “also by the general absence” to “also to the general absence”?11. (1) Please note that BSSA does not allow use of “above” or “below” when referring to other parts of an article. (2) In

this particular case, I would recommend adding a section title in place of “below” so that the sentence reads “This isimportant because one of the interpreted paleoearthquakes postdates unit 10 at the fault, as discussed in the Event 3section.” (An active link to the appropriate section would also be provided in the electronic version. Is this the correctsection title?)

12. Would it be accurate to change “We initially selected a suite of 12 samples from trench T14 to run for radiocarbondating that included charcoal from most units.” to “For carbon dating, we initially selected a suite of 12 samples fromtrench T14, each containing charcoal from most units.”?

13. Please review material from Table 2 title that has been moved to footnotes and to column heads for accuracy.14. Would C-14 be better written in standard isotope form for chemistry (as 14C where the 14 is a superscript), or does the

field of geology write it differently?15. Reimer et al. 2004 is not in the References; please provide complete information.16. Please provide the error for “(minimum of 3.0 +/ m)”.17. Added “(10.2-cm)” to be more consistent with use of metric system for other measurements. (Please note that the

metric measurement must be the unit in parentheses in order to correspond to the figures, which use inches.)18. Would it be correct to resequence the channels to “B, C, and D”?19. Add “(the oldest)” here to clarify what is meant by the “third event back”? Please suggest alternative clarification if

needed.20. Please specify the correct unit for the age range (years?).21. I tentatively changed the sentence from “One last piece of age evidence” to “Two last pieces of age evidence” because

what followed the colon appeared to be two different items. Is this correct?22. Salcedo 1640 was not cited as a reference per se anywhere in the article. Which is correct for the citation for the

sentence beginning with “Based on Salcedo’s 1640 description...”: (1) cite both the original (Salcedo, 1640) andPeraldo and Montero (1999), as is currently edited in the proof OR (2) rephrase the sentence to read “Based onthe description of the main shock by Salcedo (1640), shaking was...” and then keep subsequent the Peraldo andMontero (1999) description as currently set?

23. Is there a full name for the CASA network? All I could find on the Internet was station CASA, which is part of theLongValley network in California.


24. Is there updated status for Rockwell et al. 2010 for the References section? We need to include either (1) a volumeAND issue number or (2) a doi number if full publication information is not yet available. (It is cited several timeswithin the article.)

25. Is “Center for Accelerator Mass Spectrometry at the Lawrence Livermore National Laboratory” the correct definitionfor CAMS and LLNL? If not, please define.

26. Please provide (1) the URL where Bronk-Ramsay’s OxCal program can be accessed and (2) the month AND yearwhen you last accessed that URL for your paper.

27. Cowan 1999: Please provide total page count for report.28. Cowan et al. 1998: (1) Please include the page numbers on which the map is found OR the total page count for the

entire report. (2) Please explain the inclusion of the date “September 30, 2005.” (If this is the date when a URL wasaccessed, then please provide the complete URL. Perhaps delete as typo from two references down?)

29. ECI 2005: Page counts should include figures and appendices in this case. Please provide new total page count30. For all ECI reports, I modified the phrase “unpublished consulting report for the...” to “consulting report prepared for

the...” because consulting reports are not expected to be published; however, are these files accessible to the public ifthey were to be requested? If so, please include URL or other contact information. If not, then we will need to movethe information to the Data and Resources section with a note that indicates the information is not available to thepublic.

31. ECI 2006: Please provide new total page count that includes figures and appendices.32. ECI 2007: Please provide new total page count.33. ECI 2008: Please provide new total page count.34. Ramsey 2000: Please add month AND year when you last accessed this URL for your paper.35. (Note: Author query 24 has already asked for updated status.)36. Salcedo 1640: Because this appears to be a section within a book, a page range is required instead of a total page

count. Please provide page range.37. Westman 1947 as translator: Please provide (1) publisher name, (2) city, state/country where publisher is located, (3)

total page count.38. Technos, Inc. 2006: As with the ECI reports, is this report available to the general public if requested? If so, please

include URL or other contact information. If not, then wewill need to move the information to the Data and Resourcessection with a note that indicates the information is not available to the public.

39. URS Inc. is not not cited anywhere in your paper. Delete reference? If not: (1) please indicate where citation(s) shouldbe inserted in the article. (2) Is this report available to the general public if requested? If so, please include URL orother contact information. If not, then we will need to move the information to the Data and Resources section with anote that indicates the information is not available to the public.


Neotectonics and Paleoseismology of the Limón and Pedro...

Documents

Transcript of Neotectonics and Paleoseismology of the Limón and Pedro...